Visual intelligent processing method for electronic information data

Abstract

The invention relates to the technical field of industrial data analysis, in particular to an electronic information data visualization intelligent processing method which comprises the following steps: collecting data of a temperature sensor, a pressure sensor and a position sensor of a factory, and performing data cleaning and format standardization to obtain standardized sensor data. According to the method, data from a temperature sensor, a pressure sensor and a position sensor are collected in real time, data cleaning and format standardization are implemented, accurate fusion of data from different sources is achieved, the problems of heterogeneous and insufficient collaboration of multi-source data are solved, and comprehensiveness and consistency of data analysis are improved; intelligent definition and detection of an event mode are implemented based on fusion data, complex association of temperature anomaly and pressure mutation is captured in real time, potential association between data is deeply mined, the source of anomaly occurrence is quickly positioned, and the accuracy and reliability of data analysis are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial data analysis, and in particular to a method for intelligently processing electronic information data visualization. Background Art

[0002] The field of industrial data analysis technology mainly involves technical methods and systems for collecting, processing, analyzing and visualizing various data generated in the industrial production process, including data acquisition, data cleaning, feature extraction, intelligent analysis and modeling, prediction optimization, and interactive visualization.

[0003] In actual applications, existing technologies separately handle the collection, processing, analysis and display of industrial data, and lack effective deep integration between the various links, resulting in frequent data islands; event detection and data association analysis are mostly single-dimensional monitoring, which makes it difficult to accurately capture abnormal events with multiple factors intertwined under complex working conditions, resulting in frequent false alarms or missed alarms, increasing the difficulty of troubleshooting. Therefore, improvements are needed. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent processing method for electronic information data visualization.

[0005] In order to achieve the above object, the present invention adopts the following technical solution, a method for intelligent processing of electronic information data visualization, comprising the following steps:

[0006] Collect data from factory temperature sensors, pressure sensors, and position sensors, perform data cleaning and format standardization, and obtain standardized sensor data; use data fusion to integrate the standardized sensor data into a unified data framework to generate a fused sensor data framework;

[0007] Based on the fusion sensor data framework, define and implement data stream event patterns, including temperature exceeding a threshold and pressure mutation, perform event detection and association analysis, and obtain preliminary event association results; analyze the preliminary event association results through a graph theory algorithm to display the relationship between the data and generate a data relationship display diagram;

[0008] Using the data relationship display diagram, the data path is tracked and optimized and analyzed to identify key transmission nodes and potential data flow bottlenecks, and obtain key path analysis results; the data processing flow is adjusted according to the key path analysis results, the data transmission and processing efficiency are optimized, and an optimized data processing flow is generated;

[0009] The optimized data processing flow is applied to real-time data monitoring, and combined with augmented reality, the data is visualized in real time on the operator's visual interface.

[0010] Preferably, the step of acquiring the standardized sensor data is: collecting raw data from the temperature sensor, pressure sensor and position sensor of the factory, performing error checking and outlier screening on each sensor data, deleting the erroneous data and outliers, and obtaining preliminary cleaned data;

[0011] Based on the preliminary cleaned data, adjusting the data resolution and unifying the timestamp format to obtain format-standardized data;

[0012] The data is standardized according to the format, and data quality is checked, including verifying data integrity and consistency, to obtain standardized sensor data.

[0013] Preferably, the acquisition step of the fusion sensor data framework is: based on the standardized sensor data, all sensor data are aligned according to timestamps, the data are clustered to form a unified data structure, and preliminary integrated data is obtained;

[0014] Based on the preliminary integrated data, filtering the noise in the data, using statistics to identify and correct outliers in the data, and obtaining enhanced integrated data;

[0015] Based on the enhanced integrated data, a fused sensor data framework is obtained by combining data sources, timestamp accuracy and sensor types.

[0016] Preferably, the step of obtaining the preliminary event association result is: based on the fusion sensor data framework, extracting the temperature measurement data sequence of the temperature sensor and the pressure measurement data sequence of the pressure sensor, arranging them in the time sequence of data acquisition, calculating the time increment of the temperature measurement data sequence, and calculating the instantaneous change rate of the pressure measurement data sequence, to obtain the preliminary event data stream;

[0017] According to the preliminary event data stream, the data mutation intensity index is calculated, and the calculation formula is:

[0018]

[0019] in, is the data mutation intensity indicator, is the time increment of the temperature measurement data series, is the instantaneous rate of change of the pressure measurement data sequence, is the cross-change rate of the temperature measurement data sequence and the pressure measurement data sequence, is the smoothed rate of change of the temperature measurement data series, is the cumulative change difference of the pressure measurement data sequence;

[0020] Based on the data mutation intensity index, the data stream event pattern is defined and implemented, the change threshold of the temperature measurement data sequence and the mutation standard of the pressure measurement data sequence are set, the data mutation intensity index is classified and event pattern matching is performed, the event types that meet the conditions are identified, and the preliminary event association results are obtained.

[0021] Preferably, the step of obtaining the data relationship display diagram is: based on the preliminary event association results, extracting each event node and connection relationship, constructing an event interaction matrix, and removing isolated event nodes to obtain an event relationship matrix;

[0022] According to the event relationship matrix, the event centrality index is calculated, and the calculation formula is:

[0023]

[0024] in, is the event centrality index, For event nodes With event nodes The frequency of interaction between For Node and The number of responses between For Node and The number of new connections between For Node Total interaction time, For Node Number of projects involved, is the number of event nodes in the current network;

[0025] Based on the event centrality index, an event association topology is constructed through graph theory, a node distribution diagram is drawn, and through the connectivity analysis between nodes, the key event flow relationship is screened to form a data relationship display diagram.

[0026] Preferably, the step of obtaining the critical path analysis result is: calculating the critical path optimization index according to the data relationship display diagram, and the calculation formula is:

[0027]

[0028] in, is the critical path optimization index, For Node Traffic data, For Node The shortest path length to the key node, For Node The buffer size, For Node processing speed, For Node The number of connections, For Node The data load, For Node The volatility of traffic, is the number of nodes on the path;

[0029] Based on the critical path optimization index, potential bottlenecks and key transmission nodes in data transmission are identified, and which nodes and connections affect data flow efficiency are analyzed to obtain critical path analysis results.

[0030] Preferably, the step of obtaining the optimized data processing flow is: based on the key path analysis result, extracting the inter-node delay, bandwidth usage and data packet loss rate of the data flow to obtain the data path parameters to be optimized;

[0031] According to the data path parameters to be optimized, the path optimization priority is calculated, and the calculation formula is:

[0032]

[0033] in, Optimize the priority for the path, For path The bandwidth occupancy ratio, For path The packet loss rate, For path The average transmission time, For path The delay, For path The jitter value of

[0034] Based on the path optimization priority, the transmission strategy of the data stream is adjusted, the traffic load is redistributed, and an optimized data processing flow is generated.

[0035] Preferably, the optimized data processing flow is applied to real-time data monitoring, and combined with augmented reality, the specific steps of visually presenting the data in real time on the operator's visual interface are:

[0036] According to the optimized data processing flow, the transmission rate, the frequency of abnormal events and the data integrity detection results are selected to generate an input configuration for real-time data monitoring;

[0037] The input configuration of the real-time data monitoring is converted into a dynamic chart and a digital display of performance indicators to generate a visual data flow;

[0038] The visualized data stream is dynamically projected into the operator's augmented reality field of view to display the data view.

[0039] Compared with the prior art, the advantages and positive effects of the present invention are:

[0040] The present invention collects data from temperature sensors, pressure sensors and position sensors in real time, implements data cleaning and format standardization, realizes accurate fusion of data from different sources, solves the problems of heterogeneity and lack of synergy of multi-source data, and improves the comprehensiveness and consistency of data analysis; implements intelligent definition and detection of event patterns based on fused data, captures the complex associations of temperature anomalies and pressure mutations in real time, deeply explores the potential associations between data, quickly locates the root causes of anomalies, and improves the accuracy and reliability of data analysis; combines graph theory algorithms to deeply analyze the implicit relationships between data, intuitively presents the internal associations and mutual influence paths of multidimensional data in the form of data relationship diagrams, realizes the visualization of the relationships between data, avoids blind spots and misjudgments in the data analysis process, and improves data understanding efficiency; at the same time, through data path tracking and optimization analysis, actively identifies key transmission nodes and bottleneck positions in the data flow, timely optimizes data processing procedures and data transmission strategies, and achieves a dual improvement in data processing efficiency and real-time monitoring performance; and uses augmented reality technology to realize real-time visualization of optimized data processing procedures, so that operators can intuitively grasp data dynamics, quickly respond to changes in the production environment, and improve the agility and decision-making accuracy of real-time monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0043] See also Figure 1 The present invention provides a technical solution, a method for intelligent processing of electronic information data visualization, comprising the following steps:

[0044] Collect data from factory temperature sensors, pressure sensors, and position sensors, perform data cleaning and format standardization, and obtain standardized sensor data; use data fusion to integrate standardized sensor data into a unified data framework to generate a fused sensor data framework;

[0045] Based on the fusion sensor data framework, define and implement data stream event patterns, including temperature exceeding thresholds and pressure mutations, conduct event detection and correlation analysis, and obtain preliminary event correlation results; analyze the preliminary event correlation results through graph theory algorithms to show the relationship between data and generate data relationship display diagrams;

[0046] Use the data relationship display diagram to track and optimize the data path, identify key transmission nodes and potential data flow bottlenecks, and obtain the key path analysis results; adjust the data processing flow according to the key path analysis results, optimize data transmission and processing efficiency, and generate an optimized data processing flow;

[0047] The optimized data processing flow is applied to real-time data monitoring, and combined with augmented reality, the data is visualized in real time on the operator's visual interface.

[0048] The steps for obtaining standardized sensor data are as follows: collect the raw data of temperature sensors, pressure sensors, and position sensors from the factory, perform error checking and outlier screening on each sensor data, delete the erroneous data and outliers, and obtain preliminary cleaned data;

[0049] Based on the preliminary cleansing data, adjust the data resolution and unify the timestamp format to obtain standardized data;

[0050] The data is standardized according to the format and data quality is checked, including verification of data integrity and consistency, to obtain standardized sensor data.

[0051] Specifically, based on the information sources provided by the temperature sensors, pressure sensors, and position sensors deployed in the factory area, the measurement range and working mode of each sensor data are disassembled from the overview information of the factory production line equipment. For example, the range of the temperature sensor is between to and the range can be obtained through the equipment specifications. Then, the collected original records of temperature, pressure, and position are compared with the pre-established valid range one by one. For example, the temperature value is compared to the interval, the pressure value is compared to the interval, and the position data is compared to the interval. If it is found to be out of range, it is marked as a candidate error point, and then the statistical error threshold set based on experience and the actual monitoring scenario is used to determine the error point. These candidate error points are checked, and the statistical error threshold can be preliminarily confirmed by estimating the temperature mean and standard deviation and then taking a range of three times the standard deviation. For example, if the temperature value at a certain moment exceeds or is lower than, it is considered abnormal. The pressure and position data are calculated in the same way, and compared again with the abnormal conditions recorded in the factory's historical operation log. If it is confirmed that there is a measurement abnormality, the corresponding record is deleted. If there is an obvious jump but it is not outside the allowable range, it is regarded as a suspicious point and the stability of the adjacent moments is continued to be calculated. It is eliminated only after it is determined that it has exceeded the threshold set by experience for multiple consecutive times. Finally, all the sensor measurement items after elimination are summarized to obtain preliminary cleaning data.

[0052] According to the preliminary cleaning data obtained above, first determine a unified sampling period and resolution. For example, resample and record the temperature, pressure and position data at a time interval of 1 second. Before performing this resolution adjustment, it is necessary to select a suitable sampling period based on the production process beat and sensor update frequency given in the factory document. If it is found that the refresh rate of a certain type of sensor is 0.5 seconds or shorter, the resolution can be set to 0.5 seconds without ignoring small fluctuations. At this time, when unifying the timestamp format, it is necessary to synchronize these resolution differences to continuously use UTC time or generate standardized time records by comparing with the factory's internal clock benchmark. In this process, if it is detected that some sensors have missing timestamps or jumps exceeding a fixed time span, the time error upper limit provided in the device description of each sensor is used as a reference, and a maximum correctable range is set according to the error upper limit. For example, when a device description recommends a maximum allowable offset of 0.2 seconds, if the detected offset does not exceed 0.2 seconds, linear interpolation compensation is performed. If it exceeds, the record is marked as abnormal and discarded. Finally, through this resolution unification and timestamp alignment operation, format standardized data is obtained.

[0053] According to the format standardized data obtained above, data integrity verification and consistency check are performed in sequence. First, the number of records is compared point by point with the theoretical number of sampling times per second. If it is found that it is lower than a certain ratio threshold of the number of records that should be collected, such as 10% as an empirical threshold, it is determined to be insufficient in integrity and a missing record mark is added to the system. At the same time, the time series of temperature, pressure and position data are matched to check whether each moment is arranged in sequence and there is no cross-overlapping. If multiple different measurement values ​​are detected to be repeated at the same moment, the most reliable group is selected by comparing the measurement habits under the same conditions in the factory's historical operation log or by the similarity of the values ​​of the previous and next moments, or the data are merged according to the stable output error range of the sensor. For example, if a pressure sensor gives two values ​​at the same time, and the sensor device document states that its error range is, the record closest to the trend of the previous and next moments is retained, and the other item is marked as a duplicate value and deleted. After all duplicate or conflicting records are cleared, standardized sensor data that passes the integrity and consistency check is obtained.

[0054] The steps for obtaining the fusion sensor data framework are as follows: based on the standardized sensor data, all sensor data are aligned according to the timestamp, the data are clustered to form a unified data structure, and the preliminary integrated data is obtained;

[0055] Based on the preliminary integrated data, the noise in the data is filtered, and the outliers in the data are identified and corrected using statistics to obtain enhanced integrated data;

[0056] Based on the enhanced integrated data, the fused sensor data framework is obtained by combining the data source, timestamp accuracy and sensor type.

[0057] Specifically, based on the standardized sensor data obtained above, the temperature, pressure and position records at the same time or within 0.1 seconds are first regarded as the same time slice according to the timestamp. If some records are found to be beyond the alignment range, they are regarded as independent time slices or adjusted according to the recommended errors in the device instructions. Next, the aligned records are clustered. For example, if the distance-based grouping method is used, the Euclidean distance between the sensor readings in the same time slice can be calculated first. If the distance is less than the empirically set threshold, , then these sensor readings are clustered into one category, where The choice of can be determined by analyzing the stable range of temperature and pressure in the previous plant operation and combining it with the on-site location distribution. For example, in the test phase, the sensor readings are collected within a week and the average interval is calculated. As an initial reference, when the temperature and pressure dispersion within some clusters is found to be too large after clustering, the threshold is then corrected to 0.03 or less. After the sensor readings in each time slice are divided into several groups according to similarity, the temperature, pressure and position records in each group are taken as a comprehensive record and the results are recorded in a unified data structure. If the temperature or pressure difference between a group of readings exceeds the floating limit calculated in advance from the historical stable range, for example, the temperature is allowed to float no more than , the pressure is allowed to float no more than , then the group is subdivided or marked as possibly having cross-regional phenomena, and finally the clustering information of all time slices is integrated to obtain preliminary integrated data.

[0058] According to the preliminary integrated data obtained above, noise filtering is performed on the records in turn. If the temperature or pressure in a record continuously jumps in a short time interval and deviates from the average value of other records in the same category by more than three times the standard deviation, it is regarded as noise interference. The three times the standard deviation here can be calculated by the standard deviation of the historical readings of this category. And let the threshold be To determine, if the temperature at a certain time satisfy It is marked as a noise candidate point. The same judgment is made for pressure and position. The records that have been determined to be noise can be replaced by interpolation or the median of similar records. Then, statistical identification is performed on the replaced data to check whether there is still a large fluctuation at that moment. If it still exceeds the above-mentioned empirical threshold, it is corrected again. In this process, manual comparison and inspection will be carried out on the records with continuous abnormalities. The manual inspection threshold can be determined by the longest continuous abnormal time period in the historical curve. For example, if a sensor has been observed to have abnormal values ​​for 3 seconds in the past operation, 3 seconds will be regarded as a suspicious interval for additional confirmation. After completing the above correction, the corrected data at each moment will be summarized and associated according to the previous grouping structure. If a group still cannot maintain a relatively close distribution with the records in the same group after compensation, it will be split into a new group and verified again. After all records have completed noise filtering and abnormality correction, a new integration result is formed to obtain enhanced integrated data.

[0059] According to the enhanced integrated data obtained above, the corresponding index mapping is established in combination with the data source, timestamp accuracy and sensor type. For example, the temperature readings are distinguished according to whether they come from high-temperature zone equipment or normal-temperature zone equipment and their timestamp accuracy is marked as 0.1 second or 0.5 second. The pressure readings are distinguished according to the maximum pressure range into intervals of 0MPa to 2MPa and 2MPa to 4MPa. The position readings are broken down into different production line sections according to the factory layout. Then, each record is marked with a source label and an accuracy label in a tabular form. When reading the record, the corresponding accuracy type will be automatically matched according to the source and a secondary check will be performed. If it is found that the timestamp accuracy does not match the set value, the alignment correction amount will be recalculated according to the error range in the corresponding sensor description. If multiple records from the same source are repeatedly identified, the one with the highest stability will be extracted as the final record. When all the data are summarized and mapped in this way, the temperature, pressure and position data items are included in the unified index system, so that subsequent retrieval can be quickly located by timestamp or sensor type, and finally the fusion sensor data framework is obtained on this basis.

[0060] The steps for obtaining the preliminary event association results are as follows: based on the fusion sensor data framework, extract the temperature measurement data sequence of the temperature sensor and the pressure measurement data sequence of the pressure sensor, and arrange them in the time sequence of data acquisition, calculate the time increment of the temperature measurement data sequence, and calculate the instantaneous change rate of the pressure measurement data sequence, and obtain the preliminary event data stream;

[0061] According to the preliminary event data flow, the data mutation intensity index is calculated using the following formula:

[0062]

[0063] in, is the data mutation intensity indicator, is the time increment of the temperature measurement data series, is the instantaneous rate of change of the pressure measurement data sequence, is the cross-change rate of the temperature measurement data sequence and the pressure measurement data sequence, is the smoothed rate of change of the temperature measurement data series, is the cumulative change difference of the pressure measurement data sequence;

[0064] Based on the data mutation intensity index, the data stream event pattern is defined and implemented, the change threshold of the temperature measurement data sequence and the mutation standard of the pressure measurement data sequence are set, the data mutation intensity index is classified and event pattern matched, the event types that meet the conditions are identified, and the preliminary event association results are obtained.

[0065] Specifically, according to the fusion sensor data framework obtained above, first extract the temperature and pressure related entries in its records, gradually read the temperature measurement data sequence and pressure measurement data sequence corresponding to each moment, and sort all moments according to the time standard used in the industrial field. If the time difference between some measurement data is less than 0.02 seconds, they are regarded as the same instantaneous record and marked as the same entry. When the temperature measurement data sequence is in place, subtract the recording time between each consecutive entry to obtain the time increment of the temperature measurement data sequence. For example, when extracting, if the timestamp of the previous entry is 152.20 The timestamp of the current entry is 152.23 seconds, so its time increment is recorded as 0.03 seconds. When the field equipment runs faster and the sensor sampling frequency is high, smaller time increment values ​​may also appear. These smaller intervals are also recorded. In the subsequent judgment process, it will be confirmed whether there is an overly tight or abnormal distribution. For the pressure measurement data sequence, it is necessary to extract the numerical difference between adjacent entries and divide it by the corresponding time increment to calculate the instantaneous rate of change. For example, in the adjacent entries, the pressure value rises from 1.20MPa to 1.25MPa, and the time difference between the two is 0.05 seconds. The instantaneous rate of change can be calculated as MPa / second. If the pressure changes dramatically in a short period of time and exceeds the floating threshold set in the historical record, for example, if it exceeds 0.5MPa / second, the record will be marked as an abnormal candidate. The threshold can be set by comparing the factory's pressure equipment instructions with the monitoring records of the past 100 hours. For example, the maximum normal pressure change rate obtained in the past 100 hours is about 0.4MPa / second, so 0.5MPa / second is selected as an empirical threshold. After completing the time series calculation and instantaneous change rate analysis of all temperature and pressure entries, the corresponding relationship between the temperature time increment and the instantaneous change rate of pressure at each moment is obtained, and finally integrated into a complete record sequence to obtain a preliminary event data stream.

[0066] The benefit of the formula is that by incorporating the product of the time increment and the instantaneous rate of change of pressure into the square root operation and combining the cross-rate of change of temperature and pressure, as well as the deviation between the smoothed rate of change and the cumulative change difference, the degree of mutual influence between temperature and pressure in a short period of time can be quantified. This quantification method covers the comprehensive contribution of time resolution characteristics and differences in different dimensions in multi-dimensional monitoring data, and can provide a more detailed evaluation basis for subsequent anomaly detection and event identification in actual industrial scenarios.

[0067] The parameter acquisition step is to read the temperature measurement data sequence records one by one from the preliminary event data stream obtained above, and make a difference through the timestamps between adjacent entries. If the timestamp of the previous record is The timestamp of the next record is ,but When industrial field sensors collect data at a sampling frequency of 1 second, most The interval is about 1 second. In some cases, if the sensor has a high-frequency acquisition requirement, the interval may be 0.2 seconds or less. For example, in a certain test, three sensors on site sent 5 measurement data per second. After arranging, we can get If some time intervals are obviously deviated from the normal range, the record can be corrected according to the maximum time error calibrated in the factory equipment description. The following example shows the acquisition method. In a two-hour monitoring cycle, the sensor records a total of 720 temperature data, each corresponding to a timestamp. If the timestamps of the 250th record and the 249th record differ by 0.22 seconds, it can be recorded as Second.

[0068] The parameter acquisition step is to calculate the instantaneous change rate one by one in the pressure measurement data sequence in chronological order. If the pressure value of the previous record is , the currently recorded pressure value is ,but ,in Corresponding to the temperature time interval that coincides with the time of the pressure record, if the pressure change range of the industrial site is between 0MPa and 4MPa, the pressure values ​​at different times can be read and the pressure values ​​at the previous moment can be subtracted from them, and then divided by the corresponding , get the specific value of each instantaneous change rate. For example, if in a certain monitoring, the pressure of the 120th record and the 119th record are 1.80MPa and 1.75MPa respectively, and the time interval between the two is is 1 second, then MPa / second.

[0069] The steps for obtaining the parameters are as follows: the cross-changes between the temperature measurement data sequence and the pressure measurement data sequence in the same period need to be reflected. The temperature change and pressure change in each time period can be multiplied by a position correlation coefficient. If the position sensor record shows that the device position remains unchanged during this period, the position correlation coefficient can be 1. If the position switches, 0.8 or other empirical values ​​can be used. The specific quantification method can be based on the formula: ,in is the position correlation coefficient, which needs to be calculated based on the previously obtained position data and the classification of the field area. For example, in the equipment installation document, it is recorded that the change correlation strength between the temperature and pressure in area A of the production line is 0.85, and that in area B is 0.9. By querying the position data at that moment, it is determined that the record is located in area B, so , assuming (here "for example" replaces "assuming") that the temperature changes from becomes , the pressure changes from 1.65MPa to 1.68MPa, then .

[0070] The steps to obtain the parameters are as follows: the smooth change rate of the temperature measurement data sequence needs to be estimated. The average slope of the temperature change in each time window can be taken, and then weighted according to the volatility. If in the past 10 consecutive samples, 9 temperature sampling points have increased continuously, and the increment amplitude is within If the smoothing rate of change is within If the sequence is still in the same area, it is multiplied by a small range coefficient. For example, the small range coefficient can be between 0.7 and 0.9. The specific value can be obtained by statistically analyzing the fluctuation characteristics of temperature sampling in different working intervals. When the statistical results show that the average increment of temperature sampling for 10 consecutive times is If the temperature difference is within 0.8, you can select 0.8 as the small range coefficient. For example, in one monitoring, the total temperature difference of the past 10 samplings is , and the total time interval between the 10 sampling points is 10 seconds, then the basic smoothing value is , then multiply by 0.8 .

[0071] The parameter acquisition step is to characterize the cumulative change difference of the pressure measurement data sequence, which can be accumulated within a monitoring cycle. Specifically, the difference between each record and the initial record can be accumulated, and then compared with some smaller time periods in the intermediate process. If the cumulative amount exceeds a certain threshold, it means that the pressure has a more obvious superimposed fluctuation. The threshold can be determined in combination with the maximum working range of the equipment. For example, the normal pressure difference accumulation indicated in the factory manual is about 1.0MPa, so 1.2MPa can be selected as the threshold when analyzing the field data. For example, in a 12-hour monitoring, the average pressure value per hour is recorded and the difference relative to the initial time is accumulated in turn. If the total accumulated amount reaches 0.6MPa at the 6th hour and 1.3MPa at the 12th hour, then Exceeding the threshold of 1.2MPa, the specific calculation method is to record every hour minus If the result is greater than 1.2MPa, then Mark it as 1.3 MPa and record it.

[0072] Substituting the parameters into the formula, the calculated value is 0.06497. The result shows that during this time period, the interaction between the temperature time increment and the instantaneous rate of change of pressure is relatively small, and no significant temperature and pressure coupling mutation occurs. If E is greater than 1.0, it indicates an obvious mutation. If E is between 0.1 and 1.0, it is in the general range of variation. When it is less than 0.1, it can be regarded as stable. It can be managed in stages according to the on-site conditions, and the event type can be further identified later.

[0073] Based on the data mutation strength index E obtained above, it is necessary to compare the temperature measurement data sequence change threshold corresponding to different event types with the pressure measurement data sequence mutation standard. If the temperature measurement data sequence increment at adjacent moments exceeds the temperature sudden change threshold recorded in the industrial manual (for example, set to per second) and the calculated result of E is higher than 1.0 at the same time, it is preliminarily identified as a concurrent event of temperature surge and pressure fluctuation. If the mutation standard of the pressure measurement data sequence is obtained from the field operation record, and a slightly higher value is taken as the evaluation benchmark after the maximum stable change rate in the past 72 hours is counted (for example, 0.5MPa / second is taken as the empirical benchmark value), when it is detected that the pressure change rate exceeds the benchmark value at any time and E is significantly greater than 0.8, it will be judged as a high-intensity pressure fluctuation event. When classifying these events, those that meet the dual conditions of temperature surge and pressure fluctuation are classified as comprehensive event types, while those that only have a single pressure mutation are classified as a single type of event. Finally, by matching the event pattern and recording the time and type of each event, the time periods of all symbol conditions are sorted out and summarized to obtain the preliminary event association results.

[0074] The steps for obtaining the data relationship display diagram are as follows: based on the preliminary event association results, extract each event node and connection relationship, construct an event interaction matrix, and remove isolated event nodes to obtain an event relationship matrix;

[0075] According to the event relationship matrix, the event centrality index is calculated using the following formula:

[0076]

[0077] in, is the event centrality index, For event nodes With event nodes The frequency of interaction between For Node and The number of responses between For Node and The number of new connections between For Node Total interaction time, For Node Number of projects participated in, is the number of event nodes in the current network;

[0078] Based on the event centrality index, the event association topology structure is constructed through graph theory, the node distribution map is drawn, and the key event flow relationship is screened through the connectivity analysis between nodes to form a data relationship display diagram.

[0079] Specifically, according to the preliminary event association results obtained above, it is necessary to retrieve all the marked event information and disassemble the association records between event identifiers and events. When sorting, first summarize each event identifier and its corresponding time interval, and read the interaction description entries between these events. For example, if the mutual influence frequency or time overlap of multiple events is recorded during the monitoring period, it is regarded as an event connection relationship and the basis for each interaction is noted in the table. If some events do not have any time overlap with other events or the interaction frequency is lower than the threshold set by experience, such as less than 2 times, they are marked as isolated event nodes for removal. The empirical threshold can be determined in combination with the statistical results of historical monitoring. For example, after analyzing the event interactions of the past 72 hours by time period, it is found that Nodes with an interaction frequency lower than 2 times have no significant reference value, so this value is set to 2 times. After all event nodes and their connection relationships are extracted, an event interaction matrix is ​​constructed for the remaining interrelated nodes. The rows and columns of the matrix correspond to event identifiers, respectively. Each unit in the matrix is ​​used to record the interaction intensity or association description item between the corresponding row event and column event. If different levels of interaction are found between events in the monitoring log, they can be recorded in layers in the same unit. For example, the number of overlapping time periods and the interaction frequency are filled in separately. When the entire matrix is ​​completed, the row and column elements between all nodes are scanned to check whether there are repeated marks or empty records. The repeated marked parts are merged. The empty records indicate that there is no effective interaction. Finally, the nodes that do not have sufficient interaction support are removed and the event relationship matrix is ​​obtained.

[0080] The benefit of the formula is that it takes into account multiple interaction factors between event nodes at the same time and constructs the denominator through the total interaction time and the number of projects in which the node participates. This can quantify the comprehensive activity and connection properties of the node in the entire network structure. When a node has a higher interaction frequency, number of new contacts, and number of responses, the numerator will show an additive increase, and the higher number of participating projects is corrected through the logarithmic term of the denominator, thereby taking into account the balance between the breadth of interaction and activity.

[0081] The steps to obtain the parameter are as follows: the parameter represents the event node With event nodes The interaction frequency between them needs to be searched in the event interaction matrix or related records obtained earlier. Node and The number of interactions between nodes within a fixed time window. If multiple records in the corresponding period show that there is a data or status association between the same two events, they are accumulated by entry. If the industrial site uses a periodic monitoring method and updates the event interaction record once every hour, the statistics can be counted every hour. and If the co-occurrence of In the specific implementation, it is necessary to collect statistical values ​​from the aforementioned 72 hours or longer period, and add up the interaction frequencies in all time windows as the final For example, in a monitoring cycle, the interaction frequencies of two event nodes are 2 times, 3 times, 1 time and 4 times within 4 hours, then the total is 2+3+1+4=10. .

[0082] The steps to obtain the parameters are to reflect the node and nodes The number of mutual responses between them needs to be further subdivided based on the statistics of interaction frequency. If two event nodes not only appear in the same period of time, but also have multiple records of mutual responses, then this type of response can be regarded as a mutual response. When recording, set a recognition interval. For example, set continuous mutual visits within 30 seconds as the same response chain. If nodes are found within 1 minute, To Node The trigger signal was sent twice and the node Also for nodes If two responses are generated, this can be regarded as four responses, and the total number of responses can be accumulated. For example, in one monitoring cycle, the number of responses of two event nodes is counted and the total number of responses is 12. .

[0083] The steps to obtain the parameters are as follows: and nodes The number of new connections between nodes requires that the first interaction in the current monitoring cycle be recorded as a new connection, and repeated interactions are no longer counted as new connections. When collecting relevant records, an initial identifier can be set for each pair of nodes. If the node is detected for the first time, With Node If a connection is observed between nodes, it is counted as a new connection. If a connection of the same type or time period is observed again, it is considered a continuous connection and is not counted as a new connection. If the monitoring cycle is set to 12 hours on site, the number of first interactions between each pair of nodes within the 12 hours can be counted. For example, by traversing the data log within 12 hours, if a node is found With Node If there are 3 new associations, they will be recorded .

[0084] The steps to obtain the parameters are as follows: The total interaction time needs to be recorded at the node The duration of each interaction within the monitoring period is calculated, and these durations are accumulated to obtain the total interaction time. For example, in the monitoring, each event interaction record is set with a start and end timestamp. If the node If a node is active from 10:00 to 10:05 in one interaction, its duration is 5 minutes. If it is active from 11:15 to 11:20 in another interaction, an additional 5 minutes will be added. If such interactions occur multiple times, the duration will be further accumulated. For example, if a node is monitored within 24 hours, There were 6 interactions, each lasting 3 minutes, 5 minutes, 2 minutes, 4 minutes, 6 minutes, and 5 minutes, totaling 3+5+2+4+6+5=25 minutes. .

[0085] The steps to obtain the parameters are as follows: The number of participating projects can be checked and counted in the node background information or monitoring records. For multiple parallel projects at industrial sites, if the node If a node is involved in the event management of multiple production lines or multiple devices one after another or simultaneously, all these project numbers are counted in aggregate. For example, If you have participated in three production line projects and one maintenance project in the past week, you can count it as 4 projects. If a new safety inspection item is added, 1 more item will be added, forming a new summary value of 5.

[0086] Substituting the parameters into the calculation, the result is 0.0808, which shows that the node In the current monitoring network, it presents a moderate level of centrality. When AC is greater than 1.0, it indicates that its interactive participation and response are relatively intensive and the number of participating projects is relatively small. When AC is less than 0.5, it indicates that its interaction in the network is relatively limited. Therefore, when analyzing the flow of events, we can give priority to nodes with higher AC values ​​to identify active contact areas or frequently reciprocating event combinations.

[0087] Based on the event centrality index obtained above, it is necessary to create an associated topological distribution for all nodes from a graphical perspective. First, all nodes are arranged from high to low according to their centrality index. If the centrality index is greater than 1.0, the node is marked as highly active. If it is between 0.5 and 1.0, it is regarded as a normal range node. If it is lower than 0.5, it is marked as sparsely interacted. When drawing the node distribution map, a more eye-catching shape or color is set for the nodes with high centrality index, and connecting lines are drawn between the nodes. For event nodes with significant previous interaction frequency, the connecting lines can be distinguished by thicker lines. After completing the preliminary map, the connection relationship between nodes can be analyzed for connectivity according to the requirements of on-site monitoring. For example, the number of lines between nodes can be counted one by one. When the number of connecting lines exceeds a certain threshold (for example, 5), it is regarded as a close connection. When the number of connections between nodes is less than 2, it is regarded as a loose connection. Finally, the high-connectivity node group and its flow relationship are sorted out, and all high-connectivity events are arranged according to the formed flow order and recorded in the associated schematic table. Finally, the key event flow relationship is screened to form a data relationship display diagram.

[0088] The steps to obtain the critical path analysis results are: according to the data relationship display diagram, calculate the critical path optimization index, the calculation formula is:

[0089]

[0090] in, is the critical path optimization index, For Node Traffic data, For Node The shortest path length to the key node, For Node The buffer size, For Node processing speed, For Node The number of connections, For Node The data load, For Node The volatility of traffic, is the number of nodes on the path;

[0091] Based on the critical path optimization index, potential bottlenecks and key transmission nodes in data transmission are identified, and which nodes and connections affect data flow efficiency are analyzed to obtain critical path analysis results.

[0092] Specifically, the formula is beneficial in that it integrates multiple indicators such as node traffic data, shortest path length, buffer size, processing speed, number of connections, data load, and traffic volatility into the same calculation system. The impact of multiple factors on path efficiency is reflected by adding the numerator, and the denominator is normalized and corrected by the square root of the sum of traffic volatility. This allows for more clear identification of which nodes have a more significant impact on data flow efficiency on the entire path in a multi-node linkage environment. At the same time, the numerator fully covers the bandwidth carrying capacity and load pressure in actual scenarios, helping to identify potential bottlenecks in the overall industrial data transmission scenario.

[0093] The steps to obtain the parameter are: This parameter represents the node Traffic data requires nodes in industrial network environments Accumulate input and output traffic within a specified monitoring period. For example, a 12-hour monitoring period is selected to accurately record the node traffic every hour. The inbound and outbound traffic values ​​(in Mbps or MB / s) are then summarized during the monitoring period, and if necessary, duplicate entries are eliminated based on the actual transmission protocol. If there is a peer connection, duplicate statistics need to be excluded. After the final summary is obtained, a total value can be averaged to the level of each second or minute to obtain a stable value. For example, if a node is observed If a total of 432,000 MB of data is transferred within 12 hours, then 432,000 MB can be divided by (12 hours × 3,600 seconds) = 43,200 seconds. About 10MB / s.

[0094] The steps to obtain the parameter are: This parameter represents the node The shortest path length to the key node is usually counted for the topology between different nodes in the industrial network layout. The shortest path length here can be expressed by the number of hops or physical distance between nodes. The number of hops refers to the number of relay devices between nodes, and the physical distance is the sum of the physical cable or wireless transmission distance between nodes. For example, when the number of hops is used as the length measurement, when the node If the distance to the key node is 3 forwarding devices, If the actual monitoring uses distance (km) measurement, the total distance from the node to the key node can be obtained by adding up the lengths of each cable and using it as In terms of acquisition methods, the link table of each cable and switch can be sorted out in the topology planning stage, and nodes can be searched one by one in the table. All feasible paths to the key node and find the one with the least hops or the shortest distance, and then record this value as , for example, in the field survey, the node The physical transmission distance from the key node is 6 kilometers. Other paths have longer detours or more hops, so 6 kilometers is selected as .

[0095] The steps to obtain the parameter are: the parameter is a node The buffer size represents the maximum capacity of the node that can temporarily store data packets. The unit is usually MB or GB. It needs to be checked in the hardware specification of the node device or the system configuration file. You can also get the real-time available capacity based on the actual queuing test. For example, in a factory network, a core switching device may have a 512MB or 1GB buffer, while the edge device is usually only 256MB or smaller. The specific needs to be verified through testing. If in a continuous transmission stress test, the node is injected with traffic. Continue to send a large number of data packets, and record the cache usage when the node starts to lose packets, and then get the peak value of the available buffer. When it is finally counted that the node has used 490MB of cache before the packet loss, and the hardware description also indicates that 500MB is the upper limit, you can Recorded as 500MB.

[0096] The steps to obtain the parameter are: the parameter is a node The processing speed needs to be comprehensively measured by combining the CPU core frequency of the node, network forwarding capability and other indicators. The typical unit can be Mbps or MB / s, or it can be represented by the data packet processing volume (such as p / s). The specific role of the node in the industrial network depends on the node. For switching devices, you can continuously input the maximum amount of data that can be tolerated into the node during a 12-hour stress test and count its stable processing rate. For example, record that the node can continuously process 300MB / s of traffic without severe queuing. =300MB / s. In order to avoid fluctuations on site, multiple stress tests are usually performed and the average value is taken for safety discount. For example, the average value of multiple measurements can reach 350MB / s. In view of the probability of short-term over-peak causing queues, it can be recorded as 300MB / s in the parameter setting.

[0097] The steps to obtain the parameter are: the parameter is a node The number of connections, representing the node The number of directly connected nodes in the network. When each node in the factory network has a unique identifier and its upstream and downstream devices are clearly recorded in the topology table, the nodes can be counted by scanning the topology table. The number of directly connected neighbor nodes, such as nodes found in a network check Directly connected to 8 devices, including upper-level switches, switches of the same level, or other parallel nodes, then =8.

[0098] The steps to obtain the parameter are: the parameter is a node The data load usually refers to the total amount of data processing undertaken by the node during the monitoring period and the amount of data currently queued in the cache. The traffic in and out of the node and the internal cache are continuously monitored and converted into average load or instantaneous load values. For example, in a 24-hour observation, the number of nodes The average amount of data in the processing queue per second is about 200MB, and it once reached 500MB during the peak period. After referring to the hardware documentation, the weighted average of multiple measured values ​​can be used to obtain a stable value, such as =300MB.

[0099] The steps to obtain the parameter are: the parameter is a node The traffic volatility needs to be quantified, and the amplitude or rate of change of the traffic fluctuation within the monitoring period can be analyzed using standard deviation or coefficient of variation. For example, The transmission rate is recorded every minute for 12 hours and divided into 720 sampling points, which are recorded as , first calculate the average value of these sampling points , and then calculate the deviation of each sample point from the mean , and then divide the sum of squared deviations by the total number of sampling points to get the variance, and take the square root of the variance to get the standard deviation, and then use Or directly select a volatility indicator within a reasonable range, for example, if the node The average transfer rate is 200MB / s, and the standard deviation is about 40MB / s. .

[0100] Calculation process:

[0101] The first step is to calculate the numerator , for example, when the path contains nodes, and obtain the following values ​​(the units and formats must be unified in advance):

[0102]

[0103]

[0104]

[0105]

[0106] Calculate the sum of the three items for each node in turn:

[0107] Node 1:

[0108]

[0109] Node 2:

[0110]

[0111] Node 3:

[0112]

[0113] Node 4:

[0114]

[0115] Add the above results together:

[0116]

[0117] Step 2: Calculate the denominator For example, the traffic fluctuations of these four nodes are measured separately. ,but:

[0118]

[0119]

[0120] The third step is to combine the numerator and denominator to obtain the critical path optimization index. :

[0121]

[0122] The results show that when When the value exceeds 2.0, it means that the entire path still maintains a high comprehensive availability after combining multiple dimensional indicators such as traffic, cache, processing speed, number of connections, data load and volatility. If it is found in subsequent evaluations If the value is lower than 1.0, it means there is a conflict between traffic and load. You need to carefully check the buffering capacity or processing capacity of the node. There is a certain amount of room for optimization between 1.0 and 2.0. By properly adjusting the shortest path selection or the load distribution of each node, the data transmission efficiency of the industrial network can be improved and potential bottlenecks can be avoided.

[0123] Based on the critical path optimization index K obtained above, it is necessary to compare the node values ​​in all paths during the on-site scheduling of the industrial network, and cross-check the traffic data of each node with the shortest path length, cache occupancy ratio and processing rate information according to the calculation result of K. In this process, the node topology table drawn in the previous stage is first opened, and the actual traffic usage of each node is retrieved one by one, and the current used capacity of the buffer corresponding to each node is extracted. If it is found that the buffer occupancy of some nodes in multiple monitoring periods continues to exceed the field experience threshold (for example, when it reaches 400MB, it is considered a high-load interval), then the corresponding high-load period and the previous and next transmission nodes associated with the node are recorded, and they are tabulated and compared with the K value. If the K value of a node is close to or greater than 2.0, but the actual traffic is still on an upward trend, it means that the network path may be busy but has not yet entered a bottleneck. If the K value is between 1.0 and 2.0, and it is found that its processing speed fluctuates in the previous and next periods during the tabulation process, Then we can focus on following up the traffic trend and data load of the node in detail, and check whether the number of connections has reached the on-site threshold (for example, more than 10 connections will be marked as a densely connected node). Through this comparison, on-site operation and maintenance personnel can sort out the transmission lines one by one, and mark which nodes have gathered more connections or have high traffic turbulence in the topology table. Finally, these nodes are located as potential bottlenecks and the cache accumulation phenomenon that occurs in the relevant time period is focused on. For nodes with a longer shortest path length, we can consider whether to reduce the number of connection hops by replacing closer forwarding links. For nodes with excessively high data loads, we can coordinate surrounding nodes to share traffic. When multiple nodes meet high traffic fluctuations, we should analyze the peak synchronization of their mutual influence. Once it is confirmed that the peak arrives at the same time, we must record it. Finally, by summarizing the list of all key nodes and connections, we can determine which data connections are most likely to be blocked and which nodes affect data flow efficiency on the critical path, and obtain the key path analysis results.

[0124] The steps for obtaining the optimized data processing flow are as follows: based on the critical path analysis results, extract the inter-node delay, bandwidth usage, and packet loss rate of the data flow to obtain the data path parameters to be optimized;

[0125] According to the data path parameters to be optimized, the path optimization priority is calculated using the following formula:

[0126]

[0127] in, Optimize the priority for the path, For path The bandwidth occupancy ratio, For path The packet loss rate, For path The average transmission time, For path The delay, For path The jitter value of

[0128] Based on the path optimization priority, the data flow transmission strategy is adjusted, the traffic load is redistributed, and an optimized data processing flow is generated.

[0129] Specifically, based on the critical path analysis results obtained above, we first need to check the latency indicators between nodes in the actual usage scenarios of the industrial network. For example, we can use test data packets to record the round-trip latency one by one and summarize the recorded results in chronological order. If the latency between some nodes exceeds the preset maximum tolerance value, such as 8 milliseconds, it is marked as a high latency area. At the same time, bandwidth usage is collected. The bandwidth occupancy can be measured for each link in the monitoring software, and the average and peak bandwidth occupancy values ​​for the past 24 hours are compared. If the peak value has been maintained at 90% or above and the occupancy rate exceeds an empirical threshold for a period of time, such as 30 minutes, it is registered as a bandwidth-constrained link. The specific links where these nodes are located are recorded and the measured bandwidth peak values ​​are included in the list. For the packet loss rate, continuous sequence data can be sent within the same monitoring period and the proportion of packet loss to the total number of packets can be counted. If the loss rate exceeds the preset standard (for example, 1%), it is regarded as a potential congestion point. When comparing the packet loss rate of each node, When comparing the actual transmission volume of each node, it is also necessary to make a horizontal comparison based on the actual transmission volume of each node. When multiple nodes have a high packet loss rate in the same period, it is necessary to determine whether they are from the same link or an adjacent link. If a significant increase in packet loss is observed in multiple links, it is necessary to further check whether it is limited by the processing performance of the upstream device or the line quality. Finally, the latency, bandwidth occupancy and packet loss rate data of all nodes are summarized to form a comparable list of items, and these items are classified into categories to indicate whether they are main transmission nodes or auxiliary connections. At the same time, it is verified whether they correspond to the high-traffic fluctuation areas recorded previously. If some nodes have a latency significantly greater than 8 milliseconds and a packet loss of more than 1%, they are marked as key focus nodes. After all nodes and links are compared, the original topology can be used to color-code these high-latency links, high-bandwidth occupancy links, and high-packet loss links for easy viewing in a graphical interface, and the marked information is summarized into the data path parameters to be optimized to obtain the data path parameters to be optimized.

[0130] The benefit of the formula is that it incorporates several key indicators, including bandwidth utilization ratio, packet loss rate, average transmission time, delay and jitter value, into a single expression. The numerator is quantified by multiplying the difference between bandwidth and packet loss rate by the transmission time. At the same time, the denominator is normalized by the square root of the sum of delay and jitter values. This can more intuitively measure the priority order of the path in overall transmission performance. When the bandwidth utilization ratio is high and the packet loss rate is low, the overall numerator value is larger, but if there is a high delay or jitter at the same time, it will increase in the denominator. Therefore, the final comprehensive priority can achieve a balance between multiple dimensions.

[0131] The steps to obtain the parameter are: This parameter represents the path The bandwidth occupancy ratio of the path The bandwidth usage during the observation period is counted by time period and the ratio is calculated with the rated bandwidth upper limit of the path. For example, in an industrial Ethernet link, the rated bandwidth may be 100MB / s. If the average data transmission value reaches 80MB / s within two consecutive hours and the short-term peak value can reach 90MB / s, the occupancy ratio can be calculated to be close to 100MB / s within a certain period of time. In order to obtain a more stable , it is necessary to combine the statistical results of different time periods, or use the weighted average method. For example, if the bandwidth usage is counted every hour within 8 hours, the 8 values ​​obtained are , then the arithmetic mean is approximately 0.74.

[0132] The steps to obtain the parameter are: the parameter is the path The packet loss rate needs to be calculated by sending a known number of packets and counting the number of successfully received packets recorded by the receiver. If 10,000 packets are sent in a link and only 9,950 are successfully received, the loss rate is In order to ensure accuracy, the test is usually conducted in multiple time periods, while excluding extreme conditions. If several transmission tests of 10,000 packets each are conducted within 24 hours, the packet loss rates are recorded as , then these values ​​can be averaged to obtain .

[0133] The steps to obtain the parameter are: This parameter represents the path To calculate the average transmission time, it is necessary to record the start and end timestamps and make a difference each time data is transmitted, and then calculate the average value based on the multiple transmission durations in the entire monitoring period. For example, a large-scale file transfer is performed on an industrial site link. Each file size is the same, 500MB, and it is transferred 10 times in succession. The time consumed for each transfer is recorded as seconds, add these values ​​and divide by 10 to get the average seconds, which can then be recorded as .

[0134] The steps to obtain the parameter are: the parameter is the path The round-trip delay can be detected by ping or trace, or it can be recorded based on the built-in delay monitoring instrument in the industrial production environment. After collecting several one-way or round-trip delay values ​​within a period of time, they can be averaged or a relatively stable median value can be taken as the delay. For example, if a test is performed every 10 seconds for 30 minutes, a total of 180 delay values ​​are obtained. Statistics show that most of them are concentrated around 5 milliseconds, with the highest being 9 milliseconds and the lowest being 4 milliseconds. The sum of all values ​​is divided by 180 to obtain an average delay of about 5.3 milliseconds, which can be recorded as .

[0135] The steps to obtain the parameter are: the parameter is the path The jitter value refers to the fluctuation range of the delay between different moments, which can be obtained by calculating the standard deviation of the delay or the maximum and minimum delay differences. For example, after obtaining 180 delay values ​​in 30 minutes, calculate the average delay , and then calculate each sample and The difference between the maximum and minimum values ​​of the delay is calculated and then the square root is taken to get the standard deviation. Alternatively, the maximum and minimum values ​​of the delay are directly differentiated. The standard deviation method is more commonly used in industrial scenarios. If the statistical results show that the standard deviation of these delay values ​​is 1.2 milliseconds, then .

[0136] Calculation process: The first step is to calculate the molecule ,For example , , ,but

[0137]

[0138]

[0139] Step 2: Calculate the denominator ,For example millisecond, milliseconds, then

[0140]

[0141]

[0142] The third step is to combine the numerator and denominator to get :

[0143]

[0144] The results show that when When it is greater than 2.0, the bandwidth usage of the path is relatively sufficient and the packet loss rate is low, and it has a certain advantage in average transmission time. Even if there is a 5.3 millisecond delay and 1.2 millisecond jitter, it does not significantly lower the overall priority. If the value is lower than 1.0, it means that we need to pay special attention to the packet loss rate, bandwidth usage, or delay jitter of the path. A value between 1.0 and 2.0 is considered a medium level, and the site can decide whether further tuning is needed based on actual production needs.

[0145] Based on the path optimization priority obtained above, the network management system can check one by one whether there is uneven load distribution on any path or a surge in data volume during a critical period. In this process, each recorded path is marked value, and then Paths with a value greater than 2.0 are identified as high priority, which puts them in a more flexible traffic adjustment mode in the congestion prevention strategy. Paths between 1.0 and 2.0 are marked as medium priority, and the backup scheduling scheme is activated when the fixed duration (for example, 30 minutes) or the bandwidth occupancy is higher than 70%. If the value is lower than 1.0, it is considered as low priority. Check whether the current bandwidth resources are insufficiently configured or whether there is a situation where the packet loss rate is significantly higher than 0.5%. For these low-priority paths, you can re-plan the traffic distribution and transfer some high-traffic services to other paths. Higher priority paths are selected, and the feasible switching links between the paths are confirmed in combination with the topology diagram. If it is found during the analysis that some nodes are at the intersection of multiple low-priority paths, additional performance testing is required to check whether there are hardware failures or connection abnormalities. After completing the above inspections, a new data flow allocation plan is formed, and the new transmission delay and packet loss rate are recorded to verify the allocation effect. Finally, through multiple corrections and records, a traffic load distribution method that is beneficial to the factory's business needs is gradually sorted out to generate an optimized data processing flow.

[0146] The specific steps to apply the optimized data processing flow to real-time data monitoring and combine it with augmented reality to visualize the data in real time on the operator's visual interface are as follows:

[0147] According to the optimized data processing flow, the transmission rate, the frequency of abnormal events and the data integrity test results are selected to generate the input configuration for real-time data monitoring;

[0148] Through the input configuration of real-time data monitoring, it is converted into dynamic charts and digital displays of performance indicators to generate visual data flows;

[0149] Dynamically project visual data streams into the operator’s augmented reality field of view to present data views.

[0150] Specifically, according to the optimized data processing flow obtained above, first query the bandwidth allocation table and data flow allocation recorded therein, compare the average value of the transmission rate in each time period and the peak information one by one with the bandwidth upper limit set in the previous round of monitoring, if the transmission rate is obviously close to the rated limit in several time periods, it is marked as a high-load section, and the frequency of abnormal events corresponding to the time period is extracted in the on-site monitoring configuration, such as summarizing the number of equipment alarms that appear every hour in the past 24 hours, if the number of alarms in a certain period of time exceeds the empirical threshold continuously (for example, more than 5 times per hour), then the period is regarded as an abnormally frequent state, and the 5-time threshold here can be determined in combination with the alarm situation in the past week or longer time range, and then the type and time of occurrence of the abnormal alarm are finely split, such as distinguishing between sources such as temperature exceeding the limit and pressure mutation, and comparing with the previous data The flow allocation records are mapped, and the data source with the frequency of abnormal events is included after confirming that there is a high degree of correlation. The data integrity test results can refer to the integrity verification requirements previously formulated. For example, if the data loss or missing entries in a batch transmission exceeds 2% of the total number of entries, it is defined as integrity damage. When the actual comparison test statistics table is compared, if it is found that the integrity of individual time periods is damaged by more than 3%, the time period will be marked as a low reliability section. When integrating the above three indicators, it is necessary to classify them by time period or node, and by centrally summarizing the transmission rate, frequency of abnormal events and degree of integrity deviation, a set of detailed items covering timestamps, equipment numbers and event types are formed. Finally, by associating these items, the corresponding transmission rate peak, total number of alarms and integrity indicators of each time period are obtained, which are combined into the input configuration of real-time data monitoring.

[0151] According to the input configuration of real-time data monitoring obtained above, the transmission rate and the abnormal event type are first matched in a unified coordinate dimension, and then the alarm frequency data is segmented with time as the horizontal axis. If the transmission rate in a certain period of time has reached more than 70% of the set bandwidth upper limit, and the cumulative number of abnormal events is greater than the empirically set threshold (for example, more than 3 per hour), then the period is marked as the key analysis interval, and then the proportion of non-compliant items recorded in the data integrity test is taken and weighted. If the weighted proportion is greater than 1%, a hint of darkening the segment color is given in the visual presentation. After integrating these period nodes, a criss-crossing time series curve can be formed, and the time series curve is displayed on the curve. Additional superimposed dynamic numerical displays, for example, the transmission rate is displayed in MB / s on the left side of the vertical axis, and the number of alarms is displayed on the right side. When the cursor stays on a certain time point, the specific percentage of integrity detection can pop up. The digital display of performance indicators can be added with finer granularity. For example, in a 10-minute or 15-minute time step, the corresponding transmission rate fluctuation curve and event frequency bar chart are drawn respectively. After completing these combinations, different alarm types (such as sudden temperature increase and abnormal pressure) are superimposed on the chart with different signs. If the same type of alarm occurs multiple times on the same node in a short period of time, the marking density is increased. Finally, a dynamic chart and performance indicator digital display that can automatically switch screens over time are formed to generate a visual data flow.

[0152] According to the visualized data stream obtained above, it is necessary to load its chart and digital display layout on the augmented reality device actually used on site. First, read the transmission rate curve and alarm mark information corresponding to different nodes and map them to the coordinate position in the AR scene. If the display area resolution of the device is 1920×1080 pixels, the transmission rate curve chart is arranged in the upper area, the alarm indication list is arranged in the right area, and the dynamic progress bar of the integrity check is presented in the lower area. When new data is detected, the value of the corresponding time period will be refreshed. If the set bandwidth threshold (for example, 80% utilization rate) is exceeded, the corresponding node position will be highlighted in a brighter color. A text pop-up window is superimposed on the operator's visual interface to list the alarm types or counts of the current time period. If the corresponding integrity abnormality exceeds 2%, a color warning will be given in the progress bar below. In the actual projection process, by tracking the operator's head direction, when the operator focuses on a node, the transmission rate and alarm value of the node in the past period of time can be viewed at the same time. If you need to view longer historical data, you can switch to the adjacent time period by sliding gestures in the AR view, and have a deeper view of the alarm log or integrity check details. After integrating all visual information, this set of visual data streams is projected into the operator's augmented reality field of view to display the data view.

Claims

1. A method for intelligent processing of electronic information data visualization, characterized in that: The following steps are involved: Collect data from factory temperature sensors, pressure sensors, and position sensors, perform data cleaning and format standardization, and obtain standardized sensor data; use data fusion to integrate the standardized sensor data into a unified data framework to generate a fused sensor data framework; Based on the fusion sensor data framework, define and implement data stream event patterns, including temperature exceeding thresholds and pressure mutations, perform event detection and correlation analysis, and obtain preliminary event correlation results; Analyze the preliminary event association results by using a graph theory algorithm to display the relationship between the data and generate a data relationship display diagram; Using the data relationship display diagram, the data path is tracked and optimized and analyzed to identify key transmission nodes and potential data flow bottlenecks, and obtain key path analysis results; Adjust the data processing flow according to the critical path analysis results, optimize data transmission and processing efficiency, and generate an optimized data processing flow; The optimized data processing flow is applied to real-time data monitoring, and combined with augmented reality, the data is visualized in real time on the operator's visual interface.

2. The method for intelligent processing of electronic information data visualization according to claim 1, characterized in that: The steps of acquiring the standardized sensor data are as follows: collecting raw data from the temperature sensor, pressure sensor and position sensor of the factory, performing error checking and outlier screening on each sensor data, deleting the error data and outliers, and obtaining preliminary cleaned data; Based on the preliminary cleaned data, adjusting the data resolution and unifying the timestamp format to obtain format-standardized data; The data is standardized according to the format, and data quality is checked, including verifying data integrity and consistency, to obtain standardized sensor data.

3. The method for intelligent processing of electronic information data visualization according to claim 1, characterized in that: The acquisition step of the fusion sensor data framework is: based on the standardized sensor data, all sensor data are aligned according to timestamps, the data are clustered to form a unified data structure, and preliminary integrated data is obtained; Based on the preliminary integrated data, filtering the noise in the data, using statistics to identify and correct outliers in the data, and obtaining enhanced integrated data; Based on the enhanced integrated data, a fused sensor data framework is obtained by combining data sources, timestamp accuracy and sensor types.

4. The method for intelligent processing of electronic information data visualization according to claim 1, characterized in that: The step of obtaining the preliminary event association result is: based on the fusion sensor data framework, extracting the temperature measurement data sequence of the temperature sensor and the pressure measurement data sequence of the pressure sensor, arranging them in the time sequence of data acquisition, calculating the time increment of the temperature measurement data sequence, and calculating the instantaneous change rate of the pressure measurement data sequence, to obtain the preliminary event data stream; According to the preliminary event data stream, the data mutation intensity index is calculated, and the calculation formula is: in, is the data mutation intensity indicator, is the time increment of the temperature measurement data series, is the instantaneous rate of change of the pressure measurement data sequence, is the cross-change rate of the temperature measurement data sequence and the pressure measurement data sequence, is the smoothed rate of change of the temperature measurement data sequence, is the cumulative change difference of the pressure measurement data sequence; Based on the data mutation intensity index, the data stream event pattern is defined and implemented, the change threshold of the temperature measurement data sequence and the mutation standard of the pressure measurement data sequence are set, the data mutation intensity index is classified and event pattern matching is performed, the event types that meet the conditions are identified, and the preliminary event association results are obtained.

5. The method for intelligent processing of electronic information data visualization according to claim 1, characterized in that: The step of obtaining the data relationship display diagram is as follows: based on the preliminary event association results, extracting each event node and connection relationship, constructing an event interaction matrix, and removing isolated event nodes to obtain an event relationship matrix; According to the event relationship matrix, the event centrality index is calculated, and the calculation formula is: in, is the event centrality index, For event nodes With event nodes The frequency of interaction between For Node and The number of responses between For Node and The number of new connections between For Node Total interaction time, For Node Number of projects participated in, is the number of event nodes in the current network; Based on the event centrality index, an event association topology is constructed through graph theory, a node distribution diagram is drawn, and through the connectivity analysis between nodes, the key event flow relationship is screened to form a data relationship display diagram.

6. The method for intelligent processing of electronic information data visualization according to claim 1, characterized in that: The step of obtaining the critical path analysis result is: according to the data relationship display diagram, calculating the critical path optimization index, the calculation formula is: in, is the critical path optimization index, For Node Traffic data, For Node The shortest path length to the key node, For Node The buffer size, For Node processing speed, For Node The number of connections, For Node The data load, For Node The flow volatility, is the number of nodes on the path; Based on the critical path optimization index, potential bottlenecks and key transmission nodes in data transmission are identified, and which nodes and connections affect data flow efficiency are analyzed to obtain critical path analysis results.

7. The method for intelligent processing of electronic information data visualization according to claim 1, characterized in that: The step of obtaining the optimized data processing flow is: based on the key path analysis result, extracting the inter-node delay, bandwidth usage and data packet loss rate of the data flow to obtain the data path parameters to be optimized; According to the data path parameters to be optimized, the path optimization priority is calculated, and the calculation formula is: in, Optimize the priority for the path, For path The bandwidth occupancy ratio, For path The packet loss rate, For path The average transmission time, For path The delay, For path The jitter value of Based on the path optimization priority, the transmission strategy of the data stream is adjusted, the traffic load is redistributed, and an optimized data processing flow is generated.

8. The method for intelligent processing of electronic information data visualization according to claim 1, characterized in that: The specific steps of applying the optimized data processing flow to real-time data monitoring and combining it with augmented reality to visualize the data in real time on the operator's visual interface are as follows: According to the optimized data processing flow, the transmission rate, the frequency of abnormal events and the data integrity detection results are selected to generate an input configuration for real-time data monitoring; The input configuration of the real-time data monitoring is converted into a dynamic chart and a digital display of performance indicators to generate a visual data flow; The visualized data stream is dynamically projected into the operator's augmented reality field of view to display the data view.

Images