Networking radar-based observation quality monitoring method

Abstract

The invention discloses an observation quality monitoring method based on networking radars, relates to the cross technical field of radar technology and computer data processing, and aims to solve the problems that in the prior art, log burying points are determined depending on artificial experience, key logs are prone to being lacked, and redundancy is caused. The method comprises the steps of collecting and preprocessing multi-source historical data; defining a fault type and a log dimension set; designing an association degree scoring rule and setting a threshold value, and calculating scores to construct an association matrix; dividing necessary / selectable / useless logs, and generating a burying point list; the embedded system automatically calls necessary logs when guiding burying point configuration and faults; and regularly acquiring new data to update the matrix and the list. According to the method, manual dependence is reduced, accurate point burying is realized, the troubleshooting efficiency is improved, and dynamic operation and maintenance requirements are met.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of radar technology and computer data processing, specifically a method for monitoring observation quality based on networked radar. Background Technology

[0002] In networked radar observation quality monitoring systems, log data collection is a crucial link in supporting fault diagnosis and ensuring monitoring accuracy, requiring effective data support for different types of radar faults. However, current technologies lack the technical means to establish a correspondence between "networked radar fault types" and "log dimensions," making it impossible to clearly define which log dimensions are needed for efficient troubleshooting of a particular type of fault. The content of the log data collection points can only be determined based on manual experience. This leads to either the absence of key log dimensions in practical applications, affecting fault diagnosis efficiency, or the presence of redundant log dimensions, increasing the system's storage and data processing burden, ultimately limiting the operational efficiency of the networked radar observation quality monitoring system.

[0003] For example, when a networked radar malfunctions and affects the accuracy of the observation data, the operator cannot know which log dimension the malfunction is associated with. They can only retrieve all the embedded logs one by one to troubleshoot, which not only consumes a lot of time, but may also cause the troubleshooting to stall because the key log dimensions are not embedded, making it impossible to restore the radar to normal observation in a timely manner.

[0004] To address the above problems, this invention proposes a method for monitoring observation quality based on networked radar. Summary of the Invention

[0005] The purpose of this invention is to provide a method for monitoring observation quality based on networked radar, so as to solve the problems raised in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for monitoring observation quality based on networked radar includes the following steps: S1. Collect and preprocess multi-source historical data of the networked radar, including fault diagnosis records and log usage data; S2. Based on the preprocessed historical data, define the set of fault types and the set of log dimensions for the networked radar; S3. Design correlation scoring rules, and set correlation thresholds based on the frequency of use of log dimensions in fault diagnosis and the degree of impact of missing logs. S4. Based on the correlation scoring rules, calculate the correlation score between each fault type and each log dimension in the fault type set, and construct the fault-log dimension correlation matrix; S5. Based on the correlation matrix and preset thresholds, the log dimensions are divided into required logs, optional logs and useless logs, and a log tracking list corresponding to each fault type is generated. S6. Embed the log point list into the network radar's point deployment system to guide the configuration of log points; S7. Integrate the correlation matrix into the fault diagnosis system of the networked radar. When the radar fails, the system automatically retrieves the required log dimensions corresponding to the fault type. S8. Regularly collect new fault data and log usage data of the network radar, and update the correlation matrix and log point list.

[0007] S1 further includes the following: S1.1: Collect fault investigation records and log usage data from the network radar's operation and maintenance system and log management system; wherein, the fault investigation records include log dimension identifiers invoked during the investigation process; and the log usage data includes fault type identifiers associated with the invoked scenario. S1.2: Standardize the format of the collected multi-source historical data and unify the dimension identification of fault type descriptions and log dimension names in the data.

[0008] S2 further includes the following: S2.1: Extract the fault type identifiers corresponding to all faults from the preprocessed multi-source historical data; assign a unique type code to each unique fault type, and organize all fault types with unique codes into a structured fault type set; S2.2: Extract all invoked log dimension identifiers from the preprocessed multi-source historical data; assign a unique dimension code to each unique log dimension, and organize all log dimensions with unique codes into a structured log dimension set.

[0009] S3 further includes the following: S3.1: In the preprocessed multi-source historical data, count the number of times a single log dimension is called during the troubleshooting process of a certain fault type, and calculate the proportion of the number of calls to the total number of log calls in all troubleshooting records for that fault. Based on the log call percentage distribution characteristics of different fault types in historical data, a preset number of continuous and non-overlapping percentage intervals are divided; and a unique and fixed score value is assigned to each interval. The percentage space and score value are set based on the log call percentage distribution characteristics of different fault types in historical data. S3.2: Predefined criteria for determining the impact of missing log dimensions on the troubleshooting process, including the number of additional operation steps required for troubleshooting when the log dimension is missing, and the possibility of the troubleshooting process being interrupted due to insufficient information; The two judgment criteria are based on preprocessed historical data and quantitative rules are set. The number of operation steps is divided into a preset number of intervals. The interruption probability is quantified by the proportion of the interruption frequency to the total frequency of corresponding fault investigation, and is divided into a preset number of intervals. Based on the combination of the range of the number of operation steps and the range of the probability of interruption, a preset number of independent impact levels are defined, and a preset score is assigned to each level; the range, level and score are set based on preprocessed historical data. S3.3: Define all combinations of fault types and log dimensions that have actually been invoked in the preprocessed historical data as effective association combinations to form an effective association scoring dataset; For each combination in the effective association score dataset, the preliminary score of its usage frequency and the preliminary score of its impact are combined and the two scores are added together to calculate the preliminary comprehensive association score; and based on the distribution correspondence between these preliminary comprehensive association scores and the log dimension call ratio and the investigation interruption ratio, the boundary features that can distinguish the log dimension into high, medium and low association groups are identified. Based on the boundary characteristics, two preset thresholds for correlation scores are set, namely threshold one and threshold two. The threshold 1 serves as the dividing line between mandatory and optional logs, indicating that log dimensions with a correlation score higher than the threshold 1 are classified as mandatory logs. The threshold two serves as the dividing line between optional logs and useless logs, indicating that log dimensions with a correlation score lower than the threshold two are classified as useless logs. Log dimensions with correlation scores between threshold one and threshold two are classified as optional logs.

[0010] S4 further includes the following: S4.1: The correlation score R between a certain fault type and a certain log dimension is calculated using a weighted summation model, as shown in the following formula: R = F × W1 + Y × W2; Where F represents the frequency score of the log dimension in the corresponding troubleshooting; Y represents the impact score of the missing log dimension on the corresponding troubleshooting process; W1 represents the weight of the frequency score and W2 represents the weight of the impact score; W1 and W2 are set according to the importance of the contribution of frequency and impact to the correlation results. S4.2: Traverse each fault type with a unique type code in the fault type set and each dimension log with a unique dimension code in the log dimension set to form all combinations of fault type-log dimension; For each fault type-log dimension combination, extract the usage frequency score F and impact score Y corresponding to this fault type from the preprocessed multi-source historical data. Substitute F, Y and the preset W1 and W2 into the correlation score calculation model to obtain the correlation score R of the combination. S4.3: Construct a two-dimensional matrix structure using the unique type code of each fault type in the fault type set as the row identifier of the association matrix and the unique dimension code of each log dimension in the log dimension set as the column identifier of the association matrix; The correlation score R corresponding to each fault type-log dimension combination is filled into the cell corresponding to the intersection of the row and column labels in the two-dimensional matrix to form a structured fault type-log dimension correlation matrix.

[0011] S5 further includes the following: S5.1: Using the unique type number of the fault type in the fault type set as the index, traverse each row of the fault type-log dimension association matrix; for each row corresponding to a single fault type, extract the unique dimension code of all log dimensions in that row and the corresponding association score, and establish the correspondence between fault type code, log dimension code and association score. S5.2: For each fault type encoding - log dimension encoding - correlation score correspondence relationship, compare the correlation score with the preset threshold one and threshold two; If the correlation score is higher than the threshold, the log dimension is determined to be a required log for the corresponding fault type. If the correlation score is between threshold one and threshold two, the log dimension is determined to be an optional log for the corresponding fault type. If the correlation score is lower than the threshold of two, the log dimension is determined to be a useless log for the corresponding fault type. S5.3: Taking a single fault type as a unit, integrate the exclusive dimension codes of all log dimensions under that fault type, the log dimensions and classification results, and define the integrated data as a log tracking list.

[0012] S6 further includes the following: S6.1: Import the structured log point list into the system, and associate and bind the log point list of the corresponding fault type with the radar node according to the node code of each radar node; S6.2: The log point deployment system outputs log point configuration guidelines to each radar node based on the associated log point list, and clarifies the mandatory configuration requirements for required log dimensions and the configuration suggestions for optional log dimensions.

[0013] S7 further includes the following: S7.1: Import the fault type-log dimension correlation matrix into the fault investigation system, deploy the fault monitoring module in the fault investigation system, collect the operating status parameters of the network radar in real time, and when the parameters exceed the preset normal range and are determined to be faults, the module automatically parses the fault characteristics and extracts the type code corresponding to the fault. S7.2: The extracted fault type code is used as a row index to locate the corresponding row data in the association matrix; the dimension codes with the category identifier as mandatory logs are filtered out to form a list of mandatory log dimensions for this fault type; S7.3: Based on the selected list of required log dimensions, send a retrieval request to the log storage module; the log storage module returns the real-time log data and associated historical data for the corresponding dimension based on the request.

[0014] S8 further includes the following: S8.1: Collect new fault data and new log usage data generated within the preset collection period periodically; S8.2: Preprocess the newly collected data and merge the preprocessed data with the original historical data. For each combination of fault type in the fault type set and each log dimension in the log dimension set, use a weighted summation model to recalculate the correlation score R of the combination, and then replace the old score of the corresponding cell in the correlation matrix with the recalculated score. S8.3: Based on the updated correlation matrix and preset thresholds, re-determine the log dimension classification results corresponding to each fault type; and integrate the new classification results to generate a new log tracking point list corresponding to each fault type, replacing the original list; The updated association matrix was re-integrated into the fault diagnosis system, and the updated log tracking list was re-embedded into the tracking deployment system. The two systems were verified to be able to identify new data normally by simulating fault triggering. It was confirmed that the fault diagnosis system could retrieve the required log dimensions in the updated matrix, and the tracking deployment system could be guided by the new list for tracking configuration.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention defines a set of fault types and a set of log dimensions, designs correlation scoring rules, and constructs a fault-log dimension correlation matrix to form a structured correspondence between fault type, log dimension, and correlation score. This eliminates the need for subjective human judgment to clarify the correlation between the two, fundamentally solving the problem of reliance on human experience and ensuring the objectivity and consistency of the correlation. 2. This invention divides logs into required / optional / useless categories based on an association matrix and preset thresholds, generates a dedicated list of tracking points for each fault type, and then guides the configuration of the embedded tracking point deployment system. This allows for the precise retention of required logs (ensuring the needs of troubleshooting), the reasonable selection of optional logs, and the elimination of useless logs (reducing storage and processing costs), thereby achieving precise log tracking and balancing troubleshooting needs with system resource consumption. Attached Figure Description

[0016] Figure 1 This is a flowchart of an observation quality monitoring method based on networked radar according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example: Figure 1 As shown, the present invention provides a technical solution. A method for monitoring observation quality based on networked radar includes the following steps: S1. Collect and preprocess multi-source historical data of the networked radar, including fault diagnosis records and log usage data; S2. Based on the preprocessed historical data, define the set of fault types and the set of log dimensions for the networked radar; S3. Design correlation scoring rules, and set correlation thresholds based on the frequency of use of log dimensions in fault diagnosis and the degree of impact of missing logs. S4. Based on the correlation scoring rules, calculate the correlation score between each fault type and each log dimension in the fault type set, and construct the fault-log dimension correlation matrix; S5. Based on the correlation matrix and preset thresholds, the log dimensions are divided into required logs, optional logs and useless logs, and a log tracking list corresponding to each fault type is generated. S6. Embed the log point list into the network radar's point deployment system to guide the configuration of log points; S7. Integrate the correlation matrix into the fault diagnosis system of the networked radar. When the radar fails, the system automatically retrieves the required log dimensions corresponding to the fault type. S8. Regularly collect new fault data and log usage data of the network radar, and update the correlation matrix and log point list.

[0019] S1 further includes the following: S1.1: Collect fault investigation records and log usage data from the network radar's operation and maintenance system and log management system; wherein, the fault investigation records include log dimension identifiers invoked during the investigation process; and the log usage data includes fault type identifiers associated with the invoked scenario. S1.2: Standardize the format of the collected multi-source historical data, unify the storage format of fault investigation records, log usage statistics and operation and maintenance experience calibration data into a preset structured data format; and unify the dimension identification of fault type descriptions and log dimension names in the data. Then, based on the preset invalid data judgment rules, the historical data after the unified format is filtered to remove data with missing fields, duplicate records, and data with logical contradictions. Among them, data with missing fields refers to data that does not contain fault type identifiers and log dimension identifiers; duplicate records refer to data with completely consistent fault information and log usage information; and logically contradictory data refers to data where the fault occurrence time does not match the log collection time.

[0020] S2 further includes the following: S2.1: Define the set of fault types for the networked radar: Extract the fault type identifiers corresponding to all faults from the preprocessed historical data; then merge the fault type descriptions with the same meaning but different names, and assign a unique type code to each unique fault type, and organize all fault types with unique codes into a structured set of fault types. S2.2: Define the log dimension set of the network radar: Extract all called log dimension identifiers from the preprocessed historical data; then remove duplicate log dimensions with the same record parameter type and the same collection method; then assign a unique dimension code to each unique log dimension; finally, organize all log dimensions with unique codes into a structured log dimension set.

[0021] S3 further includes the following: S3.1: Scoring criteria for the frequency of use of design log dimensions: In the preprocessed historical data, count the number of times a single log dimension is called during the troubleshooting process of a certain fault type, and then calculate the proportion of the number of calls to the total number of log calls in all troubleshooting records of that fault. Based on the log call percentage distribution characteristics of different fault types in historical data, a preset number of continuous and non-overlapping percentage intervals are divided; and a unique and fixed score value is assigned to each interval. The percentage space and score value are set based on the log call percentage distribution characteristics of different fault types in historical data. S3.2: Design scoring rules for the impact of missing log dimensions: Define the criteria for judging the impact of missing log dimensions on the troubleshooting process, including the number of additional operation steps required for troubleshooting when the log dimension is missing, and the possibility of the troubleshooting process being interrupted due to insufficient information; and set quantitative rules for the two judgment criteria based on preprocessed historical data. The number of operation steps is divided into a preset number of intervals; the interruption probability is quantified by the proportion of the interruption frequency to the total frequency of corresponding fault investigations, and divided into a preset number of intervals; then, based on the combination relationship between the intervals of the number of operation steps and the intervals of the interruption probability, a preset number of independent impact levels are divided, and a preset score value is assigned to each level; the intervals, levels, and score values ​​are set based on preprocessed historical data. S3.3: Define all combinations of fault types and log dimensions that have actually been invoked in the preprocessed historical data as valid association combinations, exclude invalid combinations that have not been invoked, and form a valid association scoring dataset; For each combination in the effective association score dataset, the preliminary score of its usage frequency and the preliminary score of its impact are combined and the two scores are added together to calculate the preliminary comprehensive association score; and based on the distribution correspondence between these preliminary comprehensive association scores and the log dimension call ratio and the investigation interruption ratio, the boundary features that can distinguish the log dimension into high, medium and low association groups are identified. Based on the identified boundary features, two clear pre-set thresholds for correlation scores are set, namely threshold one and threshold two. Among them, threshold one serves as the dividing line between mandatory and optional logs, indicating that log dimensions with a correlation score higher than threshold one are classified as mandatory logs; Threshold 2 serves as the dividing line between optional and useless logs, indicating that log dimensions with a correlation score below threshold 2 are classified as useless logs. Log dimensions with correlation scores between threshold one and threshold two are classified as optional logs.

[0022] S4 further includes the following: S4.1: The correlation score R between a certain fault type and a certain log dimension is calculated using a weighted summation model, as shown in the following formula: R = F × W1 + Y × W2; Where F represents the frequency score of the log dimension in the corresponding troubleshooting; Y represents the impact score of the missing log dimension on the corresponding troubleshooting process; W1 represents the weight of the frequency score and W2 represents the weight of the impact score; W1 and W2 are set according to the importance of the contribution of frequency and impact to the correlation results. S4.2: Iterate through each fault type with a unique type code in the fault type set, and simultaneously iterate through each dimension log with a unique dimension code in the log dimension set, forming all combinations of fault type-log dimension; then, for each combination, extract the usage frequency score F and the impact score Y of the log dimension corresponding to this fault type from the preprocessed data, substitute F, Y and the preset W1, W2 into the correlation score calculation model to obtain the correlation score R of the combination, and then repeat the above steps so that each combination corresponds to a unique correlation score; S4.3: Construct a two-dimensional matrix structure using the unique type code of each fault type in the fault type set as the row identifier of the association matrix and the unique dimension code of each log dimension in the log dimension set as the column identifier of the association matrix; Fill the cell in the matrix corresponding to the intersection of the row and column labels of each fault type-log dimension combination with the correlation score R for each combination. After filling all cells, a structured fault type-log dimension correlation matrix is ​​formed.

[0023] S5 further includes the following: S5.1: Using the unique type number of the fault type in the fault type set as the index, traverse each row of the fault type-log dimension association matrix; for each row corresponding to a single fault type, extract the unique dimension code of all log dimensions in that row and the corresponding association score, and establish the correspondence between fault type code, log dimension code and association score. S5.2: For each fault type code - log dimension code - correlation score correspondence relationship, compare the correlation score with the preset threshold one and threshold two in turn; If the correlation score is higher than the threshold, the log dimension is determined to be a required log for the corresponding fault type. If the correlation score is between threshold one and threshold two, the log dimension is determined to be an optional log for the corresponding fault type. If the correlation score is lower than the threshold of two, the log dimension is determined to be a useless log for the corresponding fault type. Each log dimension corresponds to only one classification result, and the judgment process needs to record the correlation score of the classification basis and the comparison result of the threshold; S5.3: Taking a single fault type as the unit, integrate the exclusive dimension codes of all log dimensions under that fault type, log dimensions and classification results, integrate the data according to the preset structured format, and then define the integrated data as a log tracking point list.

[0024] S6 further includes the following: S6.1: Import the structured log tracking list into the system through the preset data import interface of the tracking deployment system. During the import process, the log tracking list corresponding to the fault type is associated and bound to the radar node according to the node code of each radar node. S6.2: The event tracking deployment system outputs event tracking configuration guidelines to each radar node based on the associated log event tracking list, and clarifies the mandatory configuration requirements for required log dimensions and the configuration suggestions for optional log dimensions; after the configuration is completed, the system automatically compares the actual event tracking configuration of the radar node with the list content.

[0025] S7 further includes the following: S7.1: Import the fault type-log dimension correlation matrix into the fault investigation system, deploy the fault monitoring module in the fault investigation system, collect the operating status parameters of the network radar in real time, and when the parameters exceed the preset normal range and are determined to be faults, the module automatically parses the fault characteristics, matches the preset fault type feature library in the system, and extracts the type code corresponding to the fault. S7.2: The system uses the extracted fault type code as the row index to locate the corresponding row data in the integrated association matrix; then it iterates through the classification results of all log dimensions for that row, filters out the dimension codes whose classification identifier is a required log, and forms a list of required log dimensions for that fault type. S7.3: The system sends a retrieval request to the log storage module based on the selected list of required log dimensions. The request includes the log dimension code and the time of the fault occurrence. The log storage module returns the real-time log data and associated historical data of the corresponding dimension based on the request. The system integrates the data according to the preset format and presents it in real time on the operation and maintenance interface.

[0026] S8 further includes the following: S8.1: Collect new fault data and new log usage data generated within the preset collection period periodically; S8.2: Preprocess the newly collected data and merge the preprocessed data with the original historical data. For each combination of fault type in the fault type set and each log dimension in the log dimension set, use a weighted summation model to recalculate the correlation score R of the combination, and then replace the old score of the corresponding cell in the correlation matrix with the recalculated new score. S8.3: Referring to the classification logic of S5, based on the updated association matrix and preset thresholds, re-determine the log dimension classification results corresponding to each fault type; then integrate the new classification results to generate a new log tracking point list corresponding to each fault type, replacing the original log tracking point list. Finally, the updated association matrix was re-integrated into the fault diagnosis system, the updated log tracking list was re-embedded into the tracking deployment system, and the two systems were verified to be able to identify new data normally by simulating fault triggering. It was confirmed that the fault diagnosis system could retrieve the required log dimensions in the updated matrix, and the tracking deployment system could guide the tracking configuration according to the new list.

[0027] This embodiment uses a regional meteorological network radar system as an application scenario. The system comprises five radar nodes (coded N01-N05) and is primarily used for atmospheric profile observation. During routine maintenance, faults such as signal attenuation, antenna pointing deviation, and data transmission interruptions frequently occur. The method of this invention is used to monitor the observation quality. The specific implementation process is as follows: First, we collected historical data from multiple sources for the networked radar. We collected nearly a year's worth of fault investigation records from the system's operation and maintenance system, including 200 valid records such as signal attenuation faults, antenna pointing deviation faults, and data transmission interruption faults. Each record was marked with the fault occurrence time (e.g., 2024-01-15 08:30) and the log dimension identifier called during the investigation (e.g., “RF module voltage log” identifier D01). We also collected log usage statistics for the same period from the log management system and recorded the associated fault type identifiers for each log dimension (e.g., D01 associated with F01, i.e., signal attenuation fault).

[0028] The collected multi-source data was then preprocessed: fault investigation records, log usage statistics, and operation and maintenance experience calibration data were uniformly converted into structured data in JSON format. The fault type descriptions were uniformly set as "signal attenuation fault (F01), antenna pointing deviation fault (F02), and data transmission interruption fault (F03)", and the log dimension names were uniformly set as "RF module voltage log (D01), antenna angle acquisition log (D02), and data transmission rate log (D03)". Data was filtered according to invalid data judgment rules, and 12 records with missing fault type identifiers (such as only recording "fault" without F01 / F02 / F03), 8 duplicate records with fault information completely consistent with log usage information, and 5 logically contradictory records with fault occurrence time (2024-02-20 14:00) and log collection time (2024-02-19 14:00) did not match. Finally, 175 valid historical data were obtained.

[0029] Based on the preprocessed valid data, a set of fault types is first defined: the type identifier and feature description corresponding to all faults are extracted. The feature description of F01 (signal attenuation fault) is "affects the strength of radar observation data, and the operating current of the RF module deviates from the normal range (10-12A) when the fault occurs", F02 (antenna pointing deviation fault) is "causes the observation angle to shift, and the antenna azimuth angle error exceeds ±0.5° when the fault occurs", and F03 (data transmission interruption fault) is "observation data cannot be uploaded, and the transmission rate is lower than 10Mbps when the fault occurs". No synonymous fault types with different names were found, and exclusive type codes were directly assigned to F01, F02, and F03, which were organized into a structured fault type set {(F01, signal attenuation fault, ...), (F02, antenna pointing deviation fault, ...), (F03, data transmission interruption fault, ...)}.

[0030] Redefine the log dimension set: Extract the identifiers and information attributes of all called log dimensions. The information attribute of D01 (RF module voltage log) is "Records the real-time voltage value of the RF module, collected every 10 seconds". The information attribute of D02 (Antenna angle acquisition log) is "Records the antenna azimuth and elevation angles, collected every 5 seconds". The information attribute of D03 (Data transmission rate log) is "Records the data upload rate, collected every 1 second". Remove one "RF module power supply log" that is completely consistent with the parameter type and acquisition method of D01. Assign exclusive dimension codes to D01, D02, and D03, and organize them into a structured log dimension set {(D01, RF module voltage log, ...), (D02, Antenna angle acquisition log, ...), (D03, Data transmission rate log, ...)}.

[0031] For each combination of fault type and log dimension, the percentage of log dimension calls in the preprocessed data is statistically analyzed: taking F01 (signal attenuation fault) as an example, its historical troubleshooting records total 50 times, and the total number of log calls is 100 times (including D01, D02, and other log dimensions). Among them, D01 was called 42 times, with a call percentage of 42÷100=42% (30%<42%≤70%), and the frequency of use is assigned a score of F=3 points according to the rules; D02 was called 12 times, with a percentage of 12% (≤30%), and F=2 points were assigned; D03 was called 5 times, with a percentage of 5% (≤30%), and F=1 point was assigned. Similarly, in troubleshooting F02 (antenna pointing deviation fault), D02 accounted for 85% (>70%), F=5 points; D01 accounted for 25%, F=2 points; D03 accounted for 10%, F=1 point; in troubleshooting F03 (data transmission interruption fault), D03 accounted for 90% (>70%), F=5 points; D01 accounted for 15%, F=1 point; D02 accounted for 8%, F=1 point.

[0032] The impact of missing log dimensions is quantitatively assessed: When investigating F01 for missing D01 (RF module voltage log), two new operations are required: "retrieving historical current data of the RF module across systems + manual on-site inspection of RF module wiring" (1-2 new steps). The investigation interruption frequency is 15 times (out of 50 total investigations), with an interruption rate of 30% (10% < 30% ≤ 50%). The corresponding impact level is medium, and the impact score is Y = 3 points. When investigating F02 for missing D02 (antenna angle acquisition log), three new operations are required: "calibrating the antenna angle sensor + recalibrating the observation benchmark" (≥ 3 new steps). The interruption frequency is 35 times (out of 40 total investigations), with an interruption rate of 87.5% (> 50%). The level is high, and Y = 5 points. When investigating F03 for missing D03 (data transmission rate log), no new operations are added, with an interruption frequency of 2 times (out of 30 total investigations), with an interruption rate of 6.7% (≤ 10%). The level is low, and Y = 1 point. The remaining combinations are judged according to the same logic. If D01 is missing and F02 is checked, one additional step is added, the interruption rate is 20%, and Y=3 points.

[0033] Extract all valid association combinations of "fault type - log dimension" from historical data (a total of 9 groups, namely F01-D01, F01-D02, F01-D03, F02-D01, F02-D02, F02-D03, F03-D01, F03-D02, F03-D03) to form a valid association score dataset. The dataset was statistically distributed according to the "percentage of call counts + percentage of interruptions": the combination of call count > 70% and interruption count > 50% (F02-D02, F03-D03) had a preliminary correlation score (calculated as F+Y) of 5+5=10 and 5+1=6 respectively, both ≥ 4 points; the combination of 30% < call count ≤ 70% and 10% < interruption count ≤ 50% (F01-D01) had a preliminary score of 3+3=6, in the range of 2-4 points; the combination of call count ≤ 30% and interruption count ≤ 10% (F01-D03, F02-D03, F03-D01, F03-D02) had preliminary scores of 1+1=2, 1+1=2, 1+1=2, 1+1=2, all ≤ 2 points. The system identifies a score of 4 as the critical value for "call rate > 70% and interruption rate > 50%" and a score of 2 as the critical value for "call rate ≤ 30% and interruption rate ≤ 10%". Therefore, it sets threshold one to 4 points (the boundary between mandatory and optional) and threshold two to 2 points (the boundary between optional and useless).

[0034] The correlation score is calculated using a weighted summation model R = F × W1 + Y × W2. Based on the operational requirements of this radar network, the frequency of use and the degree of impact contribute equally to the correlation score; therefore, weights W1 = 0.5 and W2 = 0.5 are set. Substituting the F and Y values ​​of each combination into the calculation: F01-D01: R = 3 × 0.5 + 3 × 0.5 = 3 points; F01-D02: F = 2, Y = 2, R = 2 × 0.5 + 2 × 0.5 = 2 points; F01-D03: F = 1, Y = 1, R = 1 × 0.5 + 1 × 0.5 = 1 point; F02-D01: F = 2, Y = 3, R = 2 × 0.5 + 3 × 0.5 = 2.5 points; F02-D02: F = 5, Y=5, R=5×0.5+5×0.5=5 points; F02-D03: F=1, Y=1, R=1×0.5+1×0.5=1 point; F03-D01: F=1, Y=1, R=1×0.5+1×0.5=1 point; F03-D02: F=1, Y=1, R=1×0.5+1×0.5=1 point; F03-D03: F=5, Y=1, R=5×0.5+1×0.5=3 points.

[0035] The outer traversal is initiated by sorting the fault type codes in ascending order (F01→F02→F03), locking one fault type code at a time. Then, the inner traversal is initiated by sorting the log dimension codes in ascending order (D01→D02→D03), locking one log dimension code at a time, forming unique combinations of "fault type code - log dimension code" (e.g., F01-D01, F01-D02). For each combination, the corresponding F and Y values ​​are precisely extracted from the preprocessed historical data indexed by code (e.g., F=3, Y=3 for F01-D01), and substituted into the model above to calculate the R value, ensuring that all nine combinations receive a unique correlation score without duplication or omission.

[0036] Using fault type codes (F01, F02, F03) as row identifiers and log dimension codes (D01, D02, D03) as column identifiers, construct a 3x3 two-dimensional empty matrix. Fill the corresponding cells with the R values ​​of each combination: fill 3 in row D01, 2 in column D02, and 1 in column D03; fill 2.5 in row D01, 5 in column D02, and 1 in column D03; fill 1 in row D01, 1 in column D02, and 3 in column D03. After filling, a structured association matrix is ​​formed, where each cell value represents the degree of association between the corresponding "fault type" and "log dimension".

[0037] Using the fault type code as an index, the association matrix is ​​traversed row by row: when traversing row F01, the dimension codes and corresponding R values ​​of D01 (R=3), D02 (R=2), and D03 (R=1) are extracted to establish the correspondence between "F01-D01-3, F01-D02-2, and F01-D03-1"; when traversing row F02, "F02-D01-2.5, F02-D02-5, and F02-D03-1" are extracted; when traversing row F03, "F03-D01-1, F03-D02-1, and F03-D03-3" are extracted to ensure that all log dimensions and their R values ​​corresponding to each fault type are completely extracted.

[0038] The R value of each combination is compared with the preset thresholds (4 points for threshold one, 2 points for threshold two) in sequence: R=3 for F01-D01 (2<3<4), determined as optional log; R=2 for F01-D02 (=threshold two), determined as useless log; R=1 for F01-D03 (<threshold two), determined as useless log; R=2.5 for F02-D01 (2<2.5<4), determined as optional log; R=5 for F02-D02 (>threshold one), determined as mandatory log; R=1 for F02-D03 (<threshold two), determined as useless log; R=1 for F03-D01 (<threshold two), determined as useless log; R=1 for F03-D02 (<threshold two), determined as useless log; R=3 for F03-D03 (2<3<4), determined as optional log. Each log dimension corresponds to only one classification result, and the judgment criteria are recorded (e.g., F02-D02: R=5>4, judgment is mandatory).

[0039] Data is integrated into a log point list based on a pre-defined structured format, with each fault type as a unit. The F01 list includes: fault type code F01, fault type description "signal attenuation fault, affecting observed data strength, abnormal RF module current", log dimension code D01, log dimension name "RF module voltage log", and optional classification results; D02 (antenna angle acquisition log) classification is useless; D03 (data transmission rate log) classification is useless. The F02 list includes: F02 (antenna pointing deviation fault), D02 (antenna angle acquisition log) is required, D01 is optional, and D03 is useless. The F03 list includes: F03 (data transmission interruption fault), D03 (data transmission rate log) is optional, and D01 and D02 are useless. The list format uses a table format to ensure complete and unambiguous fields.

[0040] The generated three fault type tracking lists (F01, F02, and F03) are imported into the system in JSON format via the preset API interface of the networked radar tracking system. During the import process, the lists are associated with each radar node according to its maintenance responsibilities: nodes N01 and N02 are responsible for monitoring signal attenuation faults (F01) and antenna pointing deviation faults (F02), and are associated with the F01 and F02 lists; nodes N03, N04, and N05 are responsible for monitoring data transmission interruption faults (F03), and are associated with the F03 list. The system automatically records the association relationships to ensure that each node only obtains the list that matches its responsibilities.

[0041] Based on the associated list, the event tracking deployment system outputs event tracking configuration guidelines to each node: For nodes N01 and N02, D02 (mandatory log for F02) is explicitly required to be configured, while D01 (optional log for F01) is recommended. D03 (unused logs for F01 and F02) does not need to be configured. For nodes N03-N05, D03 (optional log for F03) is recommended, while D01 and D02 do not need to be configured. After each node completes its configuration, the system automatically compares the actual configuration with the list requirements: If D02 is not configured on N01, the system triggers an alarm prompting "Required log D02 is not configured, please add it"; if D03 is configured on N03, it prompts "Configuration meets list requirements," ensuring the completeness and accuracy of the configuration.

[0042] The constructed fault-log dimension correlation matrix is ​​imported into the networked radar fault diagnosis system. A fault monitoring module is deployed in the system to collect the operating status parameters of each radar node in real time (such as RF module voltage, antenna azimuth angle, and transmission rate). When the antenna azimuth angle error of node N02 reaches 0.8° (exceeding the normal range of ±0.5°), the monitoring module automatically analyzes the fault characteristics (azimuth angle deviation, observation angle offset), matches them with the preset fault type feature library in the system, determines the fault as F02 (antenna pointing deviation fault), and extracts the fault type code F02.

[0043] The fault diagnosis system uses the extracted F02 as the row index to locate the F02 row in the integrated association matrix. It then iterates through the classification results of all log dimensions in that row (D02 required, D01 optional, D03 useless), filtering out the log dimension code D02 with the classification identifier "required," thus forming a list of required log dimensions for the F02 fault (D02 only). The filtering process requires no manual intervention; the system automatically completes dimension identification and list generation.

[0044] Based on the selected D02 code, the system sends a data retrieval request to the log storage module. The request includes the log dimension code D02 and the fault occurrence time (2024-06-10 09:15). Upon receiving the request, the log storage module extracts historical data (including antenna azimuth angle values ​​collected every 5 seconds, e.g., 89.2° at 09:10 and 90.8° at 09:15) and real-time data (azimuth angle changes after 09:15) for D02 during the period 2024-06-10 09:00-09:30. This data is then integrated into a tabular format and displayed in real-time on the operations and maintenance interface. Operations and maintenance personnel can directly view the D02 data through the interface to quickly locate the time points and trends of antenna azimuth angle offset.

[0045] According to the preset one-month data collection cycle, new fault data and new log usage data of the network radar were collected in July 2024. The new fault data includes: a new "frequency drift fault" (F04) was added to node N05, with 15 troubleshooting records. The fault characteristics are "observed frequency deviates from the preset value by 5MHz, local oscillator is abnormal". The new log usage data includes: a new log dimension "local oscillator frequency log (D04)" was added, which was called 12 times in F04 troubleshooting, accounting for 80% of the calls. When missing, two additional steps were added, and the interruption rate was 40%.

[0046] The newly collected data is preprocessed (into JSON format, removing one record missing the F04 identifier) ​​and then merged with the existing historical data. For all "fault type - log dimension" combinations (including the newly added F04 and D04), the R value is recalculated: F04-D04: F=5 (80% > 70%), Y=3 (1-2 new steps, 40% interruption), W1=0.5, W2=0.5, R=5×0.5+3×0.5=4 points. The new R value is then filled into the correlation matrix, adding a row for F04 and a column for D04, filling the F04 row and D04 column with 4. For other newly added combinations (such as F04-D01), the R value is calculated logically and filled, completing the matrix update.

[0047] Referring to the classification logic of S5, the updated matrix and thresholds (4 points, 2 points) are used to re-determine: F04-D04 has an R=4 (=threshold 1), and is determined to be a mandatory log. A list of tracking points for F04 is generated (D04 is mandatory, other dimensions are useless), replacing the old list in the system. Subsequently, the updated association matrix is ​​re-integrated into the fault diagnosis system, and the new list is embedded in the tracking point deployment system. Through simulated F04 fault verification: the system can automatically identify the F04 code, filter D04 as a mandatory log and retrieve data. After the N05 node is configured with D04 according to the list, the system verification displays "configuration meets requirements", confirming that the updated matrix and list can normally support operation and maintenance, and realize continuous iterative optimization of the solution.

[0048] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for observing quality monitoring based on networked radar, characterized in that: The method comprises the following steps: S1, collecting multi-source historical data of the networking radar and preprocessing, the multi-source historical data comprising troubleshooting records, log usage data; S2, defining a fault type set and a log dimension set of the networking radar based on the preprocessed historical data; S3, designing a correlation score rule, setting a correlation threshold based on the usage frequency of the log dimension in troubleshooting and the missing impact degree; S4, calculating the correlation score of each fault type in the fault type set and each log dimension according to the correlation score rule, and constructing a fault-log dimension correlation matrix; S5, dividing the log dimensions into mandatory logs, optional logs and useless logs based on the correlation matrix and the preset threshold, and generating a log embedding list corresponding to each fault type; S6, embedding the log embedding list into the embedding deployment system of the networking radar to guide the log embedding configuration; S7, integrating the correlation matrix into the troubleshooting system of the networking radar, and automatically calling the mandatory log dimensions of the corresponding fault type when the radar fails; S8, periodically collecting new fault data and log usage data of the networking radar, and updating the correlation matrix and the log embedding list. 2.The network-based radar-based observation quality monitoring method according to claim 1, wherein: S1 further comprises the following contents: S1.1: Collecting troubleshooting records and log usage data from the operation and maintenance system and log management system of the networking radar; wherein the troubleshooting records contain log dimension identifiers called in the troubleshooting process; and the log usage data contains fault type identifiers associated with the calling scenario; S1.2: Standardizing the format of the collected multi-source historical data, and unifying the dimension identifiers of the fault type descriptions and log dimension names in the multi-source historical data.

3. The method of claim 1, wherein: S2 further comprises the following contents: S2.1: Extracting all fault type identifiers corresponding to the faults from the preprocessed multi-source historical data; and assigning a unique type code to each unique fault type, and arranging all fault types with unique codes into a structured fault type set; S2.2: Extracting all called log dimension identifiers from the preprocessed multi-source historical data; and assigning a unique dimension code to each unique log dimension, and arranging all log dimensions with unique codes into a structured log dimension set.

4. The method of claim 1, wherein: S3 further comprises the following contents: S3.1: In the preprocessed multi-source historical data, the number of times a single log dimension is called in the troubleshooting process of a certain fault type is counted, and the proportion of the number of times in the total number of log calls in all troubleshooting records of the fault is calculated; According to the log call proportion distribution characteristics corresponding to different fault types in the historical data, a preset number of continuous and non-overlapping proportion intervals are divided; And assign a unique and fixed score value to each interval, the proportion space and score value are set based on the log call proportion distribution characteristics corresponding to different fault types in the historical data; S3.2: Predefining the judgment standard of the influence degree of missing log dimension on troubleshooting process, including the number of additional operation steps required for troubleshooting when the log dimension is missing, and the possibility of interruption of troubleshooting process due to lack of information; The two criteria are respectively based on a quantitative rule set based on the pre-processed historical data; wherein, the number of operation steps is divided into a preset number of intervals; the interruption possibility is quantified by the ratio of the interruption frequency to the total troubleshooting frequency of the corresponding fault, and is divided into a preset number of intervals; According to the combination relationship of the operation step number interval and the interruption possibility interval, a preset number of independent influence degree levels are divided, and each level is assigned a preset score value; the intervals, levels and score values are set based on the pre-processed historical data; S3.3: All actual called fault types and log dimensions in the pre-processed historical data are combined to define valid association combinations, forming an effective association score data set; For each combination in the effective association score data set, the preliminary score of the usage frequency and the preliminary score of the influence degree are combined to calculate the preliminary score of the comprehensive association degree by adding the two scores; and according to the distribution corresponding relationship between the preliminary scores of the comprehensive association degree and the log dimension call proportion, the interruption proportion in troubleshooting, the log dimension is identified to be divided into high, medium and low association groups according to the distribution corresponding relationship between the preliminary scores of the comprehensive association degree and the log dimension call proportion, the interruption proportion in troubleshooting; According to the demarcation feature, two association degree score preset thresholds are set, which are threshold one and threshold two; The threshold one is used as the division limit of the optional log and the optional log, which represents that the log dimension with an association degree score higher than the threshold one is divided into the optional log; The threshold two is used as the division limit of the optional log and the useless log, which represents that the log dimension with an association degree score lower than the threshold two is divided into the useless log; The log dimension with an association degree score between the threshold one and the threshold two is divided into the optional log.

5. The method of claim 1, wherein: S4 further comprises the following contents: S4.1: A weighted summation model is used to calculate the association degree score R of a certain fault type and a certain log dimension, and the formula is as follows: R=F×W1+Y×W2; Wherein, F represents the usage frequency score of the log dimension in the corresponding fault troubleshooting; Y represents the influence degree score of the missing log dimension on the corresponding fault troubleshooting process; W1 represents the weight of the usage frequency score, and W2 represents the weight of the influence degree score; W1 and W2 are set according to the importance of the usage frequency and the influence degree to the contribution of the association degree result; S4.2: Each fault type with a special type code in the fault type set and each dimension log with a special dimension code in the log dimension set are traversed to form all combinations of fault types and log dimensions; For each fault type-log dimension combination, the usage frequency score F and the influence degree score Y of the log dimension corresponding to the fault type are extracted from the pre-processed multi-source historical data, F, Y and the preset W1, W2 are substituted into the association degree score calculation model, and the association degree score R of the combination is obtained; S4.3: The special type code of each fault type in the fault type set is used as the row identifier of the association matrix, and the special dimension code of each log dimension in the log dimension set is used as the column identifier of the association matrix, to construct a two-dimensional matrix structure; Fill the corresponding correlation degree score R of each fault type-log dimension combination into the cell corresponding to the intersection of the row identifier and the column identifier in the two-dimensional matrix to form a structured fault type-log dimension correlation matrix.

6. The method of claim 1, wherein: S5 further includes the following: S5.1: Traverse each row of data of the fault type-log dimension correlation matrix with the exclusive type number of the fault type in the fault type set as the index; for each single fault type corresponding to each row, extract the exclusive dimension code and the corresponding correlation degree score of all log dimensions in the row to establish a corresponding relationship of fault type code-log dimension code-correlation degree score; S5.2: For each fault type code-log dimension code-correlation degree score corresponding relationship, compare the correlation degree score with the preset threshold value one and threshold value two; If the correlation degree score is higher than the threshold value one, it is determined that the log dimension is a required log of the corresponding fault type; If the correlation degree score is between the threshold value one and the threshold value two, it is determined that the log dimension is an optional log of the corresponding fault type; If the correlation degree score is lower than the threshold value two, it is determined that the log dimension is a useless log of the corresponding fault type; S5.3: Integrate the exclusive dimension code, log dimension, and classification result of all log dimensions under a single fault type, and define the integrated data as a log point list.

7. The method of claim 1, wherein: S6 further includes the following: S6.1: Import the log point list in a structured format into the system, and according to the node code of each radar node, associate and bind the log point list corresponding to the fault type with the radar node; S6.2: The point deployment system outputs the point configuration guide to each radar node according to the associated log point list, and clearly specifies the mandatory configuration requirements of the required log dimension and the configuration suggestions of the optional log dimension.

8. The method of claim 1, wherein: S7 further includes the following: S7.1: Import the fault type-log dimension correlation matrix into the fault troubleshooting system, deploy a fault monitoring module in the fault troubleshooting system, and collect the running state parameters of the networking radar in real time; when it is determined that a parameter is out of the preset normal range and is a fault, the module automatically analyzes the fault characteristics and extracts the type code corresponding to the fault; S7.2: The extracted fault type code is used as the row index to locate the corresponding row data in the correlation matrix; filter out the dimension codes with the classification identifier as the required log to form a required log dimension list of the fault type; S7.3: According to the filtered required log dimension list, send a retrieval request to the log storage module; the log storage module returns the real-time log data and associated historical data of the corresponding dimension according to the request.

9. The method of claim 1, wherein: S8 further includes the following: S8.1: According to the preset collection period, periodically collect new fault data and new log usage data generated within the period; S8.2: Preprocess the collected new data, fuse the preprocessed new data with the original historical data, and for each combination of each fault type in the fault type set and each log dimension in the log dimension set, use a weighted summation model to recalculate the correlation degree score R of the combination, and replace the old score in the corresponding cell in the correlation matrix with the recalculated score; S8.3: Based on the updated association matrix and the preset threshold, re-determine the log dimension classification result corresponding to each fault type; and integrate the new classification result to generate a new log point list corresponding to each fault type, replacing the original list; Re-integrate the updated association matrix into the fault troubleshooting system, re-embed the updated log point list into the point deployment system, and verify whether the two systems can normally identify new data by simulating fault triggering, confirming that the fault troubleshooting system can call the selected log dimensions in the updated matrix, and the point deployment system can guide point configuration according to the new list.

Images