Operation and maintenance software self-adaptive optimization method and system based on intelligent diagnosis

Abstract

The application provides a smart diagnosis-based operation and maintenance software adaptive optimization method and system, and relates to the technical field of operation and maintenance management.The method comprises the following steps: collecting operation state data of an operation and maintenance object in real time, organizing the operation state data into a multi-dimensional data set, and constructing a data feature distribution structure based on the multi-dimensional data set; defining a benchmark analysis range in the data feature distribution structure, recursively dividing the high-dimensional benchmark analysis range into eight gu limit nodes according to the data intrinsic characteristics, dynamically adjusting the division level according to the data volume and data distribution uniformity in each gu limit, and obtaining a plurality of gu limit type feature analysis areas.The application realizes the adaptiveness and continuity of operation and maintenance diagnosis and optimization, and improves the operation and maintenance efficiency and software operation stability.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance management technology, and in particular to an adaptive optimization method and system for operation and maintenance software based on intelligent diagnosis. Background Technology

[0002] With the popularization of cloud-native architecture and microservice deployment, the objects of operation and maintenance are becoming distributed and heterogeneous, and the dimensions of runtime status data are increasing dramatically. The traditional operation and maintenance model that relies on manual threshold monitoring and fixed rule diagnosis is no longer able to cope with the dynamic and complex fault scenarios.

[0003] Taking a distributed order processing cluster as an example, during peak promotional periods, it frequently experiences failures such as delayed response of some nodes and abnormal surges in the number of database connections. Traditional operation and maintenance systems can only trigger alarms through preset thresholds, requiring operation and maintenance personnel to spend a lot of time manually comparing cross-node logs and testing parameter configurations before locating the root cause as a memory leak caused by a mismatch between JVM (Java Virtual Machine) heap memory parameters and instantaneous order volume. Moreover, after manually adjusting parameters, there is a lack of continuous monitoring and dynamic adaptation mechanisms, and similar failures occur again in a short period of time. This exposes the technical defects of traditional methods, such as a lack of high-dimensional data fine feature extraction capabilities, low efficiency due to reliance on experience in root cause location, and static and fixed optimization strategies that cannot adapt to dynamic changes in the operating environment. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an adaptive optimization method and system for operation and maintenance software based on intelligent diagnosis, so as to realize the adaptive and continuous operation and maintenance diagnosis and optimization, and improve operation and maintenance efficiency and software operation stability.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] Firstly, an adaptive optimization method for operation and maintenance software based on intelligent diagnosis, the method comprising:

[0007] Real-time collection of operational status data of maintenance objects is used to organize the operational status data into a multi-dimensional dataset. A data feature distribution structure is constructed based on the multi-dimensional dataset. The baseline analysis range is defined in the data feature distribution structure. According to the inherent characteristics of the data, the high-dimensional baseline analysis range is recursively divided into eight trigram nodes. The division level is dynamically adjusted according to the amount of data and the uniformity of data distribution within each trigram, resulting in multiple trigram-type feature analysis areas.

[0008] The data of each operating status are collected into the corresponding limit feature analysis area, and feature weights are generated based on the feature deviation and data correlation of each area. Based on the preset fault feature library and combined with the feature weights, the operating status data is analyzed in real time to obtain diagnostic results.

[0009] Based on the root cause localization information in the diagnostic results, the corresponding optimization strategy is automatically matched and executed to obtain optimization execution data;

[0010] Based on the optimized execution data, the key performance indicators of the operation and maintenance objects are continuously monitored during and after the optimization strategy is executed to obtain monitoring results; the monitoring results and diagnostic results are then subjected to time series correlation analysis to obtain optimization effect evaluation data.

[0011] Based on the optimization effect evaluation data, the current optimization effect is matched with the historical diagnostic result database to obtain the secondary diagnostic result;

[0012] Based on the deviation analysis between the secondary diagnosis results and the optimized execution data, the optimization parameters are dynamically adjusted to form a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis.

[0013] Furthermore, real-time collection of operational status data of maintenance objects is performed, and this data is organized into a multi-dimensional dataset. A data feature distribution structure is then constructed based on this dataset. Within this structure, a baseline analysis range is defined. Based on the inherent characteristics of the data, the high-dimensional baseline analysis range is recursively divided into eight quaternary nodes. The division hierarchy is dynamically adjusted according to the amount of data and the uniformity of data distribution within each quaternary node, resulting in multiple quaternary feature analysis regions, including:

[0014] By collecting real-time operational status data of the objects under maintenance and organizing the operational status data into a multi-dimensional dataset, a data feature distribution structure is constructed based on the multi-dimensional dataset.

[0015] In the data feature distribution structure, the benchmark analysis range is defined based on the overall distribution range and variance of the operational status data;

[0016] Based on the benchmark analysis scope, according to the inherent characteristics of the data, the high-dimensional benchmark analysis scope is recursively divided into eight octant nodes, where each octant node represents an orthogonal subspace initialized based on data density and covariance.

[0017] Based on the partitioned nodes, the partitioning level is dynamically adjusted according to whether the amount of data in each node exceeds a preset threshold and whether the data distribution uniformity is lower than a preset uniformity threshold. Through recursive partitioning, each node meets the requirements of data volume balance and distribution uniformity, so as to obtain multiple partitioned feature analysis areas.

[0018] Furthermore, the operational status data are aggregated into the corresponding trigram feature analysis areas, and feature weights are generated based on the feature deviation and data correlation of each area. Based on a pre-set fault feature library and combined with the feature weights, the operational status data is analyzed in real time to obtain diagnostic results, including:

[0019] Based on the obtained multiple trigram-type feature analysis regions, the data of each operating state are collected into the corresponding trigram-type feature analysis region to obtain the collection results;

[0020] Based on the aggregation results, feature weights representing the importance of each feature are obtained according to the feature deviation of the data within each quaternary feature analysis area and the data correlation between different analysis areas.

[0021] Based on feature weights and combined with a pre-set fault feature library, the collected operating status data is subjected to weighted real-time analysis to obtain the results of weighted real-time analysis.

[0022] Based on the results of weighted real-time analysis, diagnostic results of root cause localization information are obtained.

[0023] Furthermore, based on the root cause localization information in the diagnostic results, the corresponding optimization strategy is automatically matched and executed to obtain optimization execution data, including:

[0024] Based on the diagnostic results, the root cause localization information is extracted from the diagnostic results. The root cause localization information includes the fault type identifier, the scope of impact parameters, and the severity level.

[0025] Based on the extracted root cause localization information, multi-dimensional matching processing is performed in the preset optimization strategy knowledge base. The strategy category is determined according to the fault type identifier, the applicable strategy range is filtered in combination with the impact range parameter, and the strategy execution priority is determined according to the severity level to obtain the matching optimization strategy set.

[0026] Based on the optimized strategy set, strategy conflict detection and fusion processing are performed. When there are multiple optional strategies, a comprehensive evaluation is conducted based on strategy execution efficiency, resource consumption cost and historical success rate data, and the final strategy is selected from the matching optimized strategy set as the target execution strategy.

[0027] Based on the target execution strategy, a corresponding strategy execution instruction set is generated and distributed to the corresponding operation and maintenance object execution node. The execution node is controlled to perform parameter adjustment, service process restart and computing resource allocation operations according to the content of the instruction set.

[0028] Based on the distributed policy execution instruction set and execution process, policy execution status data is collected in real time and integrated into optimized execution data.

[0029] Furthermore, based on the optimized execution data, the key performance indicators of the operation and maintenance objects are continuously monitored during and after the optimization strategy is executed to obtain monitoring results; the monitoring results and diagnostic results are then subjected to time series correlation analysis to obtain optimization effect evaluation data, including:

[0030] Based on the optimized execution data, the execution nodes of the operation and maintenance objects to be monitored and their corresponding monitoring time windows are determined. Within the monitoring time window, the key performance indicators of the execution nodes of the operation and maintenance objects are sampled in real time to obtain the time series data of the key performance indicators.

[0031] Based on the time series data of key performance indicators, the magnitude of change, rate of change and duration of stable state of each key performance indicator before and after the execution of the optimization strategy are calculated to obtain the trend curve of key performance indicator change, which is used as the monitoring result.

[0032] Based on the monitoring results and combined with the diagnostic results, the trend curves of key performance indicators in the monitoring results are aligned with the root cause localization information in the diagnostic results according to the timestamps to construct a time series correlation matrix.

[0033] Based on the time series correlation matrix, we analyze the causal relationship strength, time delay response characteristics and recovery degree between the changing trends of key performance indicators and root cause localization information, and calculate the quantitative indicators of optimization effect.

[0034] Based on the quantitative indicators of optimization effect, combined with the preset effect evaluation threshold, optimization effect evaluation data is obtained, which includes optimization success indicators, effect level scores, and residual problem indicators.

[0035] Furthermore, based on the optimization effect evaluation data, the current optimization effect is matched with the historical diagnostic result database to obtain secondary diagnostic results, including:

[0036] Receive optimization effect evaluation data, extract optimization success identifiers, effect level scores and residual problem indicators from the evaluation data, and use them as the current optimization effect feature vector;

[0037] Based on the feature vector of the current optimization effect, retrieve the historical optimization case with the highest similarity to the feature vector of the current optimization effect from the preset historical diagnostic result database;

[0038] Based on the retrieved historical optimization cases, the feature vector of the current optimization effect is calculated and the feature similarity between the current optimization effect and each historical optimization case is calculated. The top N most similar historical optimization cases are selected according to the similarity ranking to obtain the historical pattern matching set.

[0039] Based on the historical pattern matching set, the correlation between the diagnostic results of each historical optimization case and the current optimization effect is analyzed. When there are historical optimization cases in the historical pattern matching set with similarity exceeding a preset threshold, the diagnostic results of the historical optimization cases are inherited as secondary diagnostic results. When there are no historical optimization cases that meet the similarity threshold, secondary diagnostic results are regenerated based on the residual problem indications in the current optimization effect evaluation data.

[0040] Furthermore, based on the deviation analysis between the secondary diagnostic results and the optimized execution data, the optimization parameters are dynamically adjusted to form a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis, including:

[0041] Receive the secondary diagnostic results, extract the root cause confirmation status, continuous evaluation of optimization effect, and potential risk warning information from the secondary diagnostic results, and obtain the secondary diagnostic result information;

[0042] Based on the extracted secondary diagnostic results and combined with the optimization execution data, the deviation between the actual optimization effect index and the expected optimization target is calculated. The temporal characteristics and impact dimensions of the deviation are analyzed, and a deviation analysis report containing the deviation magnitude, deviation type and deviation impact range is generated.

[0043] Based on the deviation analysis report, identify the key factors that cause the deviation, including improper configuration of optimization strategy parameters, dynamic changes in the execution environment, or insufficient root cause localization accuracy, and determine the type and direction of optimization parameters that need to be adjusted.

[0044] Based on the key factors of deviation, the core parameters in the optimization strategy are dynamically adjusted, including resource adjustment threshold, strategy execution frequency, parameter adjustment step size and associated strategy triggering conditions, to generate an adjusted set of optimization parameter configurations.

[0045] Based on the adjusted optimized parameter configuration set, the corresponding strategy parameters in the preset optimization strategy knowledge base are updated, and the updated strategy parameters are fed back to the optimization strategy matching stage, triggering a new round of diagnosis, optimization, and verification processes, forming a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis.

[0046] Secondly, an adaptive optimization system for operation and maintenance software based on intelligent diagnosis includes:

[0047] The acquisition module is used to collect real-time operational status data of the maintenance objects, organize the operational status data into a multi-dimensional data set, and construct a data feature distribution structure based on the multi-dimensional data set. In the data feature distribution structure, the baseline analysis range is defined, and according to the inherent characteristics of the data, the high-dimensional baseline analysis range is recursively divided into eight trigram nodes. The division level is dynamically adjusted according to the amount of data and the uniformity of data distribution within each trigram, resulting in multiple trigram-type feature analysis areas.

[0048] The diagnostic module is used to collect various operating status data into the corresponding limit feature analysis area, generate feature weights based on the feature deviation and data correlation of each area, and perform real-time analysis of the operating status data based on the preset fault feature library and the feature weights to obtain diagnostic results.

[0049] The optimization module is used to automatically match and execute corresponding optimization strategies based on the root cause localization information in the diagnostic results, and obtain optimization execution data.

[0050] The monitoring module is used to continuously monitor the changing trends of key performance indicators of the operation and maintenance objects based on the optimization execution data during and after the optimization strategy is executed, so as to obtain monitoring results; and to perform time series correlation analysis on the monitoring results and diagnostic results to obtain optimization effect evaluation data.

[0051] The matching module is used to perform pattern matching between the current optimization effect and the historical diagnostic result database based on the optimization effect evaluation data to obtain the secondary diagnostic result;

[0052] The adjustment module is used to perform deviation analysis based on the secondary diagnostic results and the optimized execution data, dynamically adjust the optimization parameters, and form a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis.

[0053] Thirdly, a computing device includes:

[0054] One or more processors;

[0055] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0056] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0057] The above-described solution of the present invention has at least the following beneficial effects:

[0058] By employing core technologies such as high-dimensional data bounded recursive partitioning, feature deviation and correlation weighted diagnosis, multi-dimensional optimization strategy matching and conflict fusion, time series correlation analysis quantitative evaluation, and secondary diagnostic closed-loop parameter adjustment, this technology effectively overcomes the technical problems of traditional operation and maintenance, including the curse of dimensionality, uneven feature partitioning, low efficiency and poor accuracy due to reliance on manual experience in high-dimensional data processing, single and conflict-prone optimization strategy matching, difficulty in quantifying execution effects, and high failure recurrence rates due to the lack of dynamic adaptation mechanisms. This achieves the technical effects of refined high-dimensional data feature extraction, automated and accurate root cause localization, efficient and low-conflict optimization strategy execution, objective and quantifiable optimization effect evaluation, and adaptive closed-loop operation and maintenance processes. This improves the operational stability and efficiency of operation and maintenance software, and reduces manual operation and maintenance costs and the risk of failure recurrence. Attached Figure Description

[0059] Figure 1This is a flowchart illustrating the adaptive optimization method for operation and maintenance software based on intelligent diagnosis provided in an embodiment of the present invention.

[0060] Figure 2 This is a schematic diagram of an adaptive optimization system for operation and maintenance software based on intelligent diagnosis, provided in an embodiment of the present invention.

[0061] Figure 3 This is a schematic diagram of a computing device. Detailed Implementation

[0062] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0063] like Figure 1 As shown, embodiments of the present invention propose an adaptive optimization method for operation and maintenance software based on intelligent diagnosis, the method comprising the following steps:

[0064] Step 1: Collect real-time operational status data of the maintenance objects, organize the operational status data into a multi-dimensional dataset, and construct a data feature distribution structure based on the multi-dimensional dataset; define the benchmark analysis range in the data feature distribution structure, and recursively divide the high-dimensional benchmark analysis range into eight quaternary nodes according to the inherent characteristics of the data. Dynamically adjust the division level according to the amount of data and the uniformity of data distribution within each quaternary node to obtain multiple quaternary feature analysis areas.

[0065] Step 2: Collect the data of each operating status into the corresponding limit feature analysis area, and generate feature weights based on the feature deviation and data correlation of each area; perform real-time analysis of the operating status data based on the preset fault feature library and in combination with the feature weights to obtain the diagnostic results;

[0066] Step 3: Based on the root cause localization information in the diagnostic results, automatically match and execute the corresponding optimization strategy to obtain optimization execution data;

[0067] Step 4: Based on the optimized execution data, continuously monitor the changing trends of key performance indicators of the operation and maintenance objects during and after the optimization strategy is executed to obtain monitoring results; perform time series correlation analysis on the monitoring results and diagnostic results to obtain optimization effect evaluation data.

[0068] Step 5: Based on the optimization effect evaluation data, perform pattern matching between the current optimization effect and the historical diagnostic result database to obtain the secondary diagnostic result;

[0069] Step 6: Based on the secondary diagnosis results and the optimized execution data, perform deviation analysis, dynamically adjust the optimization parameters, and form a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis.

[0070] In this embodiment of the invention, by employing a series of technical means, including high-dimensional data bounded recursive partitioning, feature weight-driven intelligent fault diagnosis, root cause-oriented automatic matching and execution of optimization strategies, time series correlation analysis of monitoring results and diagnostic results, historical pattern matching for secondary diagnosis, and deviation analysis-driven closed-loop parameter adjustment, the technical problems of coarse high-dimensional operating status data feature extraction, low efficiency and poor accuracy of root cause location, insufficient adaptability of optimization strategies, lack of effective evaluation of optimization effects, and lack of dynamic adaptive closed-loop mechanism in traditional operation and maintenance are effectively overcome. This achieves the technical effects of refined operation and maintenance data feature extraction, automated and accurate fault diagnosis, targeted optimization strategy execution, quantified evaluation of optimization effects, and adaptive closed-loop operation and maintenance process, thereby improving the adaptability and operational stability of operation and maintenance software to dynamic operating environments.

[0071] In a preferred embodiment of the present invention, step 1 above may include:

[0072] Step 1.1 involves collecting real-time operational status data of the maintenance objects and organizing this data into a multi-dimensional dataset. Based on this dataset, a data feature distribution structure is constructed. Specifically, this includes deploying various types of data acquisition components for distributed, heterogeneous maintenance objects, including nodes in a distributed order processing cluster and database servers. These components are adaptable to different operating environments and can capture various types of data generated during the operation of the maintenance objects in real time. This includes resource utilization-related data such as CPU usage, memory usage, disk read / write speed, and network bandwidth usage; business operation-related data such as node response time, order processing success rate, and number of database connections; and log-related data such as error logs and operation logs. The collected heterogeneous data is preprocessed. First, duplicate and invalid data are removed. Then, the units of measurement for various types of data are standardized. For example, time data in different formats are standardized to the same time standard, and data values ​​are standardized to a consistent measurement scale. The data is then classified and organized according to the collection time, indicator type, and the identification of the operation and maintenance object to form a structured multidimensional data set. This ensures that each data item in the set contains clear attribution information and time attributes. Based on this multidimensional data set, the values ​​of each data dimension are analyzed one by one, as well as the mutual influence relationship between different dimensions of data. For example, the correlation trend between memory usage data and node response time data is analyzed, and a data feature distribution structure that clearly presents the distribution location and aggregation state of all data in high-dimensional space is constructed.

[0073] Step 1.2: In the data feature distribution structure, based on the overall distribution range and variance of the operational status data, the benchmark analysis range is defined. Specifically, this includes: in the constructed data feature distribution structure, for each data dimension, statistically analyzing the values ​​of all valid data under that dimension, determining the minimum and maximum values ​​of the data in each dimension, thereby clarifying the overall distribution range of the dimension data, combining the distribution ranges of all dimensions to obtain the overall distribution range of the operational status data, performing dispersion analysis on the operational status data of each dimension, statistically analyzing the degree of deviation of the data from the average value under that dimension, and calculating the variance of the data in each dimension. The stability of the data is judged by the size of the variance; the larger the variance, the more drastic the data fluctuation and the more likely outliers are to appear. Combining the overall distribution range and variance results of the data in each dimension, the benchmark analysis range is defined. First, based on the overall distribution range of each dimension, and then referring to the dispersion reflected by the variance, those extreme outliers that deviate too far from the overall distribution are eliminated. For example, data with variances exceeding a set reasonable range are judged as extreme values ​​and excluded. Finally, a benchmark analysis range that can cover the vast majority of valid operational status data is defined.

[0074] Step 1.3: Based on the benchmark analysis scope, and according to the inherent characteristics of the data, the high-dimensional benchmark analysis scope is recursively divided into eight octagonal nodes. Each octagonal node represents an orthogonal subspace initialized based on data density and covariance. Specifically, this includes: using the defined benchmark analysis scope as the analysis boundary, deeply exploring the inherent characteristics of the data, specifically analyzing the correlation between data in each dimension, such as determining whether there is a positive or negative correlation between memory usage data and database connection count data, and simultaneously analyzing the density distribution of data in the high-dimensional space, identifying dense and sparse regions. A high-dimensional coordinate system is established with the mean of each dimension of data as the center, and then the high-dimensional benchmark analysis scope is extended along each dimension. The system splits the data along the positive and negative directions, resulting in eight initial quaternary nodes. For higher-dimensional spaces (three dimensions and above), the splitting logic is the same as for three-dimensional spaces, continuing to expand along the positive and negative directions of each new dimension to ensure that eight quaternary nodes are always formed. During the splitting process, data density statistics are performed on the spatial region corresponding to each quaternary node to calculate the number of data samples contained in a unit space. At the same time, the covariance between the data of each dimension within each node is analyzed to clarify the degree of linear correlation between dimensions. Based on the statistically obtained data density and covariance results, each quaternary node is initialized and configured to make each quaternary node an orthogonal subspace with independent data characteristics and no interference between dimensions.

[0075] Step 1.4: Based on the partitioned trigram nodes, dynamically adjust the partitioning level according to whether the data volume within each node exceeds a preset threshold and whether the data distribution uniformity is lower than a preset uniformity threshold. Through recursive partitioning, ensure that each node meets the requirements for data volume balance and distribution uniformity to obtain multiple trigram-type feature analysis regions. Specifically, this includes: performing detailed data statistical analysis on each partitioned trigram node. First, count the total number of running status data samples contained in each node. Then, analyze the distribution of data within the corresponding node space to determine whether the data is evenly distributed. For example, observe whether the data is concentrated in a local area of ​​the node space or evenly distributed throughout the entire node space. Two judgment thresholds are preset: one is a preset data volume threshold to define the reasonable number of data samples a single node can carry, and the other is a preset distribution uniformity threshold to define the reasonable distribution of data within the node space. The uniformity is assessed by comparing the actual data volume of each node with a preset data volume threshold, and simultaneously comparing the actual data distribution uniformity of each node with a preset uniformity threshold. If the actual data volume of a node exceeds the preset data volume threshold, it indicates that the data within that node is too concentrated. If the actual data distribution uniformity of a node is below the preset uniformity threshold, it indicates that the data distribution within that node is disordered, which will affect the accuracy of feature extraction. For nodes that do not meet the requirements, the octagonal boundary recursive division is performed again according to the splitting logic to generate new child nodes. Then, the data volume and uniformity of the newly generated child nodes are judged again. The above division and judgment process is repeated until the data volume of all nodes is within the preset data volume threshold range and the data distribution uniformity of all nodes meets the preset uniformity threshold requirement. Finally, multiple octagonal boundary-type feature analysis areas with balanced data volume, uniform distribution, clear boundaries, and well-defined data characteristics are formed.

[0076] In this embodiment of the invention, because a data feature distribution structure is constructed based on a multidimensional dataset, and the benchmark analysis range is defined by combining the overall distribution range and variance of the running status data, the high-dimensional benchmark range is recursively divided into eight orthogonal subspace quaternary nodes based on data density and covariance according to the inherent characteristics of the data, and the division level is dynamically adjusted according to the node data volume threshold and distribution uniformity threshold, the technical problems of lack of scientific basis for benchmark range definition, fixed and rigid partitioning method leading to the dimensionality disaster in high-dimensional data processing, and imbalance and uneven distribution of feature partition data are effectively overcome in traditional high-dimensional data feature extraction. Thus, the accurate construction of high-dimensional data feature distribution and reasonable definition of benchmark analysis range are achieved, and finally, a quaternary feature analysis area with balanced data volume, uniform distribution and clear boundaries is obtained.

[0077] In a preferred embodiment of the present invention, step 2 above may include:

[0078] Step 2.1: Based on the obtained multiple hexagram-type feature analysis regions, the operational status data are aggregated into the corresponding hexagram-type feature analysis regions to obtain the aggregation results. Specifically, this includes: first, clarifying the spatial boundary range and core data feature attributes of each hexagram-type feature analysis region. These attributes are determined based on the previously constructed high-dimensional data feature distribution structure, accurately reflecting the unique data characteristics of each region; second, performing multi-dimensional value analysis on each operational status data sample, comparing the values ​​of each dimension of the data sample with the spatial boundaries of each hexagram-type feature analysis region, and simultaneously determining whether the feature attributes of the data sample are consistent with the core feature attributes of the corresponding analysis region; third, assigning the data samples to the hexagram-type feature analysis region with the highest matching degree of spatial boundary and feature attributes according to the comparison results, ensuring that each data can be assigned to the region that best matches its characteristics; fourth, after all operational status data samples have been allocated, counting the total number of data samples aggregated in each hexagram-type feature analysis region, organizing the specific data entries in each region, clarifying the key information such as the collection time and the identification of the maintenance object to which each data belongs, forming a complete and clear aggregation result.

[0079] Step 2.2: Based on the aggregation results, according to the feature deviation of the data within each hexagram feature analysis area and the data correlation between different analysis areas, the feature weights representing the importance of each feature are obtained. Specifically, this includes: first, obtaining the standard data distribution of each hexagram feature analysis area under normal operating conditions of the maintenance object. This standard data distribution is formed through long-term accumulation of historical normal operating data and accurately reflects the data characteristics of the maintenance object when there are no faults; for each hexagram feature analysis area, all aggregated data are compared with the corresponding standard data distribution dimension by dimension, and the degree of deviation of each data point from the standard value range in each dimension is statistically analyzed. Combining the deviation of all data in this area, a comprehensive calculation is performed to obtain the feature weights representing the importance of each feature. The feature deviation of the analysis area is analyzed. The higher the deviation, the greater the possibility of data anomalies in that area. The trend of data value changes between different limit-type feature analysis areas is analyzed. For example, when the memory usage data of a certain analysis area continues to rise, observe whether the node response time data of other analysis areas also increases. This is used to determine the degree of mutual influence between data in different areas and obtain data correlation. The stronger the correlation, the more obvious the linkage between different features. The impact of feature deviation and data correlation on fault diagnosis is comprehensively considered, and the two are quantitatively evaluated. Features with high feature deviation and strong data correlation are assigned higher importance weights. Finally, feature weights that accurately represent the importance of each data feature are generated.

[0080] Step 2.3: Based on feature weights and combined with a pre-set fault feature library, perform weighted real-time analysis on the collected operational status data to obtain the results of the weighted real-time analysis. Specifically, the pre-set fault feature library contains information such as multi-dimensional feature combinations corresponding to various known faults, fault type identifiers, common fault manifestations, impact range characteristics, and historical processing records. The information is continuously accumulated and updated by sorting out past maintenance fault cases to ensure coverage of common fault types and typical characteristics. The operational status data collected in each trigram-type feature analysis area is processed in real-time by region. According to the generated feature weights, corresponding weight values ​​are assigned to different dimensions of the data features. The higher the weight value, the greater the weight of the feature in the analysis process. The weighted operational status data in each region is compared one by one with various fault feature combinations in the fault feature library. The focus is on analyzing the matching of core features with higher weights with fault features. At the same time, the degree of fit between the current data features and various fault features is judged by combining the time-series change trend of the data. The comparison results of each data sample are recorded, and the comparison results of all regions are summarized to distinguish between fault types that are completely matched, partially matched, and mismatched, forming detailed weighted real-time analysis results.

[0081] Step 2.4: Based on the results of the weighted real-time analysis, obtain the diagnostic results of the root cause localization information. Specifically, this includes: sorting out the weighted real-time analysis results, filtering out fault types corresponding to fault characteristics that meet the set matching standards, excluding irrelevant fault types with extremely low matching degrees, narrowing the fault investigation scope, and analyzing the propagation patterns of the selected fault types among distributed nodes in conjunction with the architectural characteristics of the distributed operation and maintenance objects. For example, a certain type of fault typically spreads from database nodes to application nodes. Based on the current distribution of collected data, determine the number of operation and maintenance object nodes actually affected by the fault, the specific node identifiers, and the related businesses, thus clarifying the scope of the fault's impact. By analyzing the magnitude of the characteristic deviation, the strength of data correlation, and the breadth of the impact of the fault, the degree of interference of the fault with the normal operation of the maintenance object is assessed, the severity level of the fault is classified, and minor faults, general faults, serious faults, and fatal faults are distinguished. By combining the fault type, impact range, severity level, and the core dimension with the most significant deviation of data characteristics, the initial cause of the fault is traced. For example, by the high deviation of memory-related characteristics and the strong correlation with response time characteristics, the core root cause of improper memory parameter configuration is located. Finally, a complete diagnostic result including fault type, impact range, severity level, and core root cause is formed.

[0082] In this embodiment of the invention, by first collecting the operating status data into the corresponding quaternary feature analysis area, then generating feature weights representing the importance of features based on the feature deviation of the data in each analysis area and the correlation between the data in different analysis intervals, and finally combining the collected data with a preset fault feature library for weighted real-time analysis, the technical means effectively overcome the technical problems in traditional fault diagnosis, such as failure to distinguish the importance of features after data collection, large interference from irrelevant features, and failure to consider the correlation between different feature areas, which lead to fuzzy root cause localization, low diagnostic accuracy, and poor efficiency. This achieves the effects of accurate data collection, highlighting key fault features, and reducing interference from irrelevant features, ultimately improving the accuracy and real-time performance of root cause localization.

[0083] In a preferred embodiment of the present invention, step 3 above may include:

[0084] Step 3.1: Based on the diagnostic results, extract the root cause localization information from the diagnostic results. The root cause localization information includes the fault type identifier, the scope of impact parameters, and the severity level. Specifically, this includes: first, fully parsing the obtained diagnostic results, which contain various detailed information related to the fault, and accurately separating the core information required for root cause localization; second, identifying the fault type identifier to clarify the specific category of the current fault, such as memory leak fault, connection anomaly fault, or excessive resource consumption fault, to ensure accurate identification of the fault type; third, extracting the scope of impact parameters to determine the actual number of business modules involved in the fault-affected operation and maintenance object nodes and the affected business processes, clearly understanding the scope of the fault's spread; finally, based on the assessment conclusion of the fault's impact degree in the diagnostic results, determining the severity level, which is usually divided into four levels: minor fault, general fault, severe fault, and fatal fault. Different levels correspond to the degree of interference with the normal operation of the operation and maintenance object and the emergency handling requirements. Finally, the complete root cause localization information, including the fault type identifier, scope of impact parameters, and severity level, is compiled.

[0085] Step 3.2: Based on the extracted root cause localization information, multi-dimensional matching processing is performed in the pre-set optimization strategy knowledge base. The strategy category is determined according to the fault type identifier, the applicable strategy range is filtered based on the impact range parameter, and the strategy execution priority is determined according to the severity level to obtain a set of matching optimization strategies. Specifically, the pre-set optimization strategy knowledge base is built by sorting through a large number of historical fault handling cases, industry best practices, and the experience of operation and maintenance experts. The base stores optimization strategies corresponding to various faults, and the strategies are categorized and indexed according to the fault type, impact range, and severity for easy retrieval. Based on the extracted root cause localization information, a categorized search is first performed in the optimization strategy knowledge base according to the fault type identifier. The process involves identifying the corresponding strategy categories, such as parameter adjustment strategies and service process restart strategies for memory leak faults. Then, it filters the applicable strategy scope based on the impact range parameter. If the fault only affects some distributed nodes, optimization strategies targeting global nodes are excluded, retaining only those applicable to local nodes. If the fault involves a specific business module, optimization strategies specifically targeting that module are selected to ensure a precise match between the strategy and the fault's impact range. Finally, the execution priority of the strategies is determined based on the severity level. Strategies for fatal faults have the highest priority and must be executed immediately, while strategies for minor faults have lower priority and can be executed as needed. Through multi-dimensional matching and filtering, a set of matching optimization strategies containing multiple candidate strategies is formed.

[0086] Step 3.3: Based on the optimized strategy set, perform strategy conflict detection and fusion processing. When multiple optional strategies exist, a comprehensive evaluation is conducted based on strategy execution efficiency, resource consumption cost, and historical success rate data. The final strategy is selected from the matching optimized strategy set as the target execution strategy. Specifically, this includes: analyzing all candidate strategies in the matching optimized strategy set one by one, comparing the execution conditions, operation objects, and expected effects of different strategies, and detecting whether there are mutually exclusive strategy executions. For example, if one strategy requires the service to continue running while another strategy requires restarting the service, it is determined to be a conflict. For the detected conflicting strategies, further analyze the cause of the conflict, and combine the root cause location information of the fault with the operation and maintenance. Based on the current operating status of the target strategy, conflicting strategies are integrated and adjusted, contradictory operational steps are eliminated, and complementary effective content is retained to form a conflict-free strategy combination. When there are multiple conflict-free optional strategies in the optimization strategy set, relevant data on the past execution of each strategy are collected, and the execution efficiency of each strategy (i.e., the time consumed from execution to effectiveness, resource consumption cost, i.e., the amount of CPU, memory, network, and other resources used during execution) and historical success rate (i.e., the percentage of successful cases in handling similar faults in the past) are statistically analyzed. According to the preset weight allocation rules, the execution efficiency, resource consumption cost, and historical success rate are weighted and calculated, and a comprehensive score is assigned to each optional strategy. Finally, the strategy with the highest comprehensive score is selected as the target execution strategy.

[0087] Step 3.4: Based on the target execution strategy, generate the corresponding strategy execution instruction set and distribute it to the corresponding operation and maintenance object execution nodes. Control the execution nodes to perform parameter adjustments, service process restarts, and computing resource allocation operations according to the instruction set content. Specifically, this includes: for the determined target execution strategy, breaking down the included operation steps, clarifying the execution content, execution standard, execution order, and expected results of each step; and generating a standardized strategy execution instruction set according to the specific requirements of the operation steps. Each instruction in the instruction set contains key information such as the execution object identifier, operation type, specific parameters, and execution time limits. For example, for a target execution strategy for memory leak faults, the instruction set may include instructions to adjust the initial and maximum values ​​of the JVM heap memory of a specific node, instructions to restart the corresponding application service process, and instructions to allocate additional CPU resources. By adapting to the communication mechanism of the distributed operation and maintenance environment, the policy execution instruction set is accurately distributed to the corresponding operation and maintenance object execution node, ensuring that each node can receive and identify its own corresponding execution instruction; the instruction reception status of each execution node is monitored in real time, and the distribution is re-initiated for nodes that have not successfully received the instruction, ensuring the integrity of the instruction transmission, and controlling each execution node to execute operations such as parameter adjustment, service process restart, and computing resource allocation in sequence according to the order and requirements of the instruction set, avoiding secondary failures caused by incorrect operation order or improper parameter settings during the execution process.

[0088] Step 3.5: Based on the distributed policy execution instruction set and execution process, collect policy execution status data in real time and integrate the collected policy execution status data into optimized execution data. Specifically, this includes: immediately initiating a data collection mechanism after the policy execution instruction set is distributed to continuously capture relevant data from each execution node. The collected data includes instruction reception confirmation information, instruction start execution time, execution progress of each operation step, log information generated during execution, result feedback after operation completion, and status data such as resource usage and response time of the operation and maintenance objects during execution. All collected data is categorized and organized chronologically, grouping data from the same execution node and the same operation step together, clearly identifying the instruction identifier, execution node, and time node corresponding to each piece of data. The collected data is validated to remove duplicate, invalid, and abnormal data, ensuring data accuracy and completeness. The validated valid data is then integrated according to a preset data format and classification rules to form optimized execution data containing information about the entire instruction execution process.

[0089] In this embodiment of the invention, the technical means of first extracting root cause localization information including fault type identifier, impact range parameters, and severity level, then performing multi-dimensional matching based on this information in the optimization strategy knowledge base to form an optimization strategy set, and then screening target execution strategies through conflict detection and fusion processing, combined with a comprehensive evaluation of strategy execution efficiency, resource consumption cost, and historical success rate, and finally generating an instruction set for distribution and execution and collecting and integrating optimization execution data in real time, overcomes the technical problems of traditional optimization strategy matching being single-dimensional, not considering strategy conflicts, lacking comprehensive evaluation basis for selection, and having scattered and unintegrated execution process data, resulting in poor strategy adaptability, high execution conflict rate, serious resource waste, and lack of reliable data support for subsequent optimization effect evaluation. This achieves significantly improved accuracy and adaptability of optimization strategy matching, reduced strategy execution conflict rate, reasonable control of resource consumption, and efficient and controllable execution process, while forming complete and standardized optimization execution data.

[0090] In a preferred embodiment of the present invention, step 4 above may include:

[0091] Step 4.1: Based on the optimized execution data, determine the execution nodes of the operation and maintenance objects to be monitored and their corresponding monitoring time windows. Within the monitoring time window, sample the key performance indicators of the execution nodes of the operation and maintenance objects in real time to obtain the time-series data of the key performance indicators. Specifically, this includes: first, comprehensively analyzing the optimized execution data to extract the execution node identifiers of the operation and maintenance objects that have executed the optimization strategy, clarifying the specific type of the node, ensuring that the node to be monitored accurately corresponds to the implementation object of the optimization operation, and combining the execution duration and expected effective period of the optimization strategy to define a reasonable monitoring time window. The window needs to cover the baseline data collection stage before the optimization strategy is executed, the real-time monitoring stage during the execution process, and the effect stabilization stage after the execution is completed, to ensure complete capture. To capture performance changes before and after optimization, based on the fault scenarios of the distributed order processing cluster in the background, the specific types of key performance indicators (KPIs) are determined, including node response time, database connection count, CPU utilization, memory usage, order processing success rate, disk read / write speed, and network bandwidth usage. These indicators directly reflect the operating status and optimization effect of the maintenance object. Within the defined monitoring time window, the KPIs of each node to be monitored are continuously sampled in real time according to the preset sampling frequency. The specific values ​​and corresponding timestamps of each sample are recorded to ensure the continuity and timeliness of the sampled data. All sampled data are classified and organized according to node type, indicator type, and time order to form structured KPI time-series data.

[0092] Step 4.2: Based on the time series data of key performance indicators (KPIs), calculate the magnitude of change, rate of change, and duration of stable state of each KPI before and after the implementation of the optimization strategy, and obtain the KPI change trend curve as the monitoring result. Specifically, this includes: First, extracting the time node from the KPI time series data to start the implementation of the optimization strategy, and using this as the dividing point to divide the time series data into a pre-optimization data group and a post-optimization data group; For each KPI, calculate the difference between the statistical values ​​such as the mean and median of the post-optimization data group and the corresponding statistical values ​​of the pre-optimization data group, and then combine the statistical values ​​of the pre-optimization data group to calculate the magnitude of change of the indicator. The magnitude of change directly reflects the degree of improvement of the indicator by the optimization strategy. By analyzing the slope of the indicator value change over time in the post-optimization data group, calculate the amount of change of the indicator per unit time to obtain the rate of change. The rate of change reflects how quickly the optimization strategy takes effect. Continuously track the changes in the values ​​of key performance indicators after optimization. When the indicator values ​​stabilize within the preset normal range and no longer fluctuate significantly, record the duration from when the indicator reaches the normal range to the end of the monitoring time window. This duration is the duration of the stable state. The longer the stable time, the more lasting the optimization effect. Integrate information such as the change magnitude, rate of change, and duration of the stable state of each key performance indicator before and after optimization, and plot curves in the order of timestamps to form a clear trend curve of key performance indicator change that shows the dynamic change process of the indicator. The curve is the complete monitoring result.

[0093] Step 4.3: Based on the monitoring results and combined with the diagnostic results, align the trend curves of key performance indicators in the monitoring results with the root cause localization information in the diagnostic results according to timestamps, and construct a time series correlation matrix. Specifically, this includes: first, reviewing the root cause localization information in the diagnostic results to identify key time-related information such as fault occurrence time, fault type identifier, impact range parameters, and severity level; extracting the specific timestamps corresponding to this information to ensure the accuracy of the time information; extracting the trend curves of each key performance indicator from the monitoring results; identifying the timestamp corresponding to each data point on the curve to ensure the integrity of the time dimension of the curve; and using timestamps as the basis for correlation. A baseline is established by aligning the trend curves of key performance indicators (KPIs) with root cause analysis information point by point. For example, the KPI value corresponding to the fault occurrence time stamp is associated with the fault type identifier, and the KPI change corresponding to the optimization execution start time stamp is associated with the severity level. Based on the aligned time series data, a time series correlation matrix is ​​constructed. The row dimension of the matrix consists of continuous time stamps, and the column dimension contains all the contents of KPIs and root cause analysis information. Each cell in the matrix is ​​filled with the specific value or status identifier of the corresponding column dimension under the corresponding time stamp. The matrix visually presents the correspondence between KPI changes and root cause analysis information in the time dimension.

[0094] Step 4.4: Based on the time series correlation matrix, analyze the causal relationship strength, latency response characteristics, and recovery degree between the changing trends of key performance indicators and root cause localization information, and calculate the quantitative indicators of optimization effect. Specifically, this includes: based on the constructed time series correlation matrix, analyzing the correspondence between the changing trend curve of each key performance indicator and the impact range of the fault type in the root cause localization information. For example, when the root cause is memory leak, determine the correlation between the downward trend of memory utilization rate and the root cause. By statistically analyzing the number of synchronous changes and the degree of fit between the two in the time dimension, the causal relationship strength is quantified. The higher the causal relationship strength, the closer the indicator change is to the root cause of the fault. The closer the correlation between the causes, the more effective the optimization strategy becomes. The difference between the time stamp when key performance indicators begin to show significant changes and the time stamp when the optimization strategy begins execution is extracted to obtain the latency response characteristics. The shorter the latency, the faster the optimization strategy takes effect. The stable values ​​of the optimized key performance indicators are compared with the standard values ​​under normal operating conditions. The proportion of indicators recovering to the standard values ​​is calculated, and the recovery degree is obtained by combining this with the duration of the stable state. The higher the recovery degree, the more ideal the optimization effect. By comprehensively considering the three dimensions of causal relationship strength, latency response characteristics, and recovery degree, scientific quantitative calculation rules are established to transform the evaluation results of the three dimensions into specific values, forming quantitative indicators of optimization effect.

[0095] Step 4.5: Based on the quantitative indicators of optimization effect and combined with the preset effect evaluation thresholds, obtain optimization effect evaluation data including optimization success indicators, effect level scores, and residual problem indicators. Specifically, this includes: a preset optimization effect evaluation threshold system, which is formulated based on the performance standards of the operating object during normal operation and historical optimization effect data. This system includes thresholds for change magnitude, change rate, stable state duration, and recovery degree, etc. Different key performance indicators correspond to different threshold standards. For example, the memory utilization recovery degree threshold is set to reach more than 90% of the normal range, and the node response time change magnitude threshold is set to shorten by more than 30%. The calculated quantitative indicators of optimization effect are compared one by one with the preset effect evaluation thresholds. If all the quantitative indicators of key performance indicators meet or exceed the corresponding thresholds... If the optimization is successful, it is marked as "Yes". If some indicators fail to reach the threshold or show a reverse change, the optimization is marked as "No". Based on the degree of fit between the quantitative indicators and the threshold, an effect level scoring standard is set. For example, a quantitative indicator that significantly exceeds the threshold is an excellent score, one that just meets the threshold is a qualified score, and one that does not meet the threshold is a unqualified score. The optimization effect is scored according to this standard to obtain the effect level score. The trend curves of key performance indicators and quantitative indicators are analyzed in depth. If some indicators have not returned to the normal range or there are still slight fluctuations after optimization, the unresolved problem types and corresponding indicator data are recorded in detail to form residual problem indicators. The optimization success mark effect level score and residual problem indicators are integrated to obtain complete optimization effect evaluation data.

[0096] In this embodiment of the invention, by employing techniques such as accurately determining monitoring nodes and time windows based on optimized execution data and collecting time-series data of key performance indicators, generating trend curves by calculating the magnitude, rate, and duration of indicator changes, aligning the root cause location information of monitoring and diagnostic results with timestamps to construct a time-series correlation matrix, analyzing the strength of the causal relationship, time delay response characteristics, and recovery degree between the two to calculate quantitative indicators of optimization effect, and combining preset thresholds to output evaluation data including optimization success indicators, effect level scores, and residual problem indications, the invention effectively overcomes the technical problems in traditional optimization effect evaluation, such as lack of targeted monitoring, superficial indicator analysis, failure to establish a correlation between monitoring and diagnosis, subjective and ambiguous evaluation results, and inability to identify residual problems. This achieves the technical effects of more precise optimization effect monitoring, deeper indicator analysis, quantified evaluation basis, and comprehensive evaluation results.

[0097] In a preferred embodiment of the present invention, step 5 above may include:

[0098] Step 5.1: Receive optimization effect evaluation data and extract optimization success identifiers, effect level scores, and residual problem indicators from the evaluation data as the current optimization effect feature vector. Specifically, this includes: First, receiving the complete output optimization effect evaluation data, which contains core content such as optimization success identifiers, effect level scores, and residual problem indicators. This data directly reflects the effectiveness of optimization strategies implemented for distributed order processing clusters and other operational objects. Accurately extract optimization success identifiers from the evaluation data to clarify whether the current optimization operation successfully resolved the fault or failed to achieve the expected goal, such as whether it effectively alleviated node response latency or abnormal spikes in database connection counts. Extract the effect level score, a quantitative result calculated based on changes in key performance indicators, reflecting the degree of optimization effectiveness, such as the specific score range corresponding to different levels like excellent, good, qualified, and unqualified. Extract residual problem indicators in detail to clarify unresolved issues after optimization, such as whether some nodes still have excessively long response times, occasional fluctuations in database connection counts, or abnormal JVM heap memory usage. Integrate the extracted optimization success identifiers, effect level scores, and residual problem indicators from these three dimensions to form a feature vector comprehensively representing the current optimization effect.

[0099] Step 5.2: Based on the current optimization effect feature vector, retrieve the historical optimization cases with the highest similarity to the current optimization effect feature vector from the preset historical diagnostic result library. Specifically, the preset historical diagnostic result library stores a large amount of past optimization case data for distributed heterogeneous operation and maintenance objects. Each case contains complete optimization effect feature information and corresponding diagnostic results. For example, past cases of handling memory leaks, node response delays, and other faults record the optimization success indicator, effect level score, residual problem indication, and final diagnostic conclusion at that time. Based on the constructed current optimization effect feature vector, start the retrieval process of the historical diagnostic result library. First, according to the type of the current operation and maintenance object, such as a distributed order processing cluster, and the core fault type, such as memory-related faults and database connection-related faults, filter out historical cases in the library that match the scenario to narrow the search scope. For the filtered scenario-matching cases, compare them one by one according to the three dimensions of the feature vector, focusing on the overlap of residual problem indications, the consistency of optimization success indicators, and the closeness of effect level scores. Initially retrieve a batch of historical optimization cases with high similarity to the current optimization effect feature vector.

[0100] Step 5.3: Based on the retrieved historical optimization cases, calculate the feature similarity between the current optimization effect feature vector and each historical optimization case. Rank the historical optimization cases by similarity and select the top N most similar historical optimization cases to obtain a historical pattern matching set. Specifically, this includes: analyzing each retrieved similar historical optimization case one by one, extracting the corresponding historical optimization effect feature vector from each case. This vector also includes the historical case's optimization success indicator, effect level score, and residual problem indication. Calculate the similarity between the current optimization effect feature vector and the feature vector of each historical case from three core dimensions. For the optimization success indicator dimension, a higher similarity value is assigned if they are consistent, and a lower value if they are inconsistent. For the effect level score dimension, calculate the difference between the two scores; the smaller the difference, the higher the similarity. For the residual problem indication dimension, count the number of overlapping problem types; the higher the overlap, the higher the similarity. Combine the similarity calculation results from the three dimensions to obtain the comprehensive similarity between each historical case and the current scenario. Rank all retrieved historical cases in descending order of comprehensive similarity. Based on actual operation and maintenance needs and the size of the historical case database, set a reasonable number N and select the top N historical optimization cases to form a historical pattern matching set.

[0101] Step 5.4: Based on the historical pattern matching set, analyze the correlation between the diagnostic results of each historical optimization case and the current optimization effect. When there are historical optimization cases in the historical pattern matching set with similarity exceeding a preset threshold, inherit the diagnostic results of the historical optimization cases as secondary diagnostic results. When there are no historical optimization cases that meet the similarity threshold, regenerate secondary diagnostic results based on the residual problem indications in the current optimization effect evaluation data. Specifically, this includes: conducting in-depth analysis of each historical optimization case in the historical pattern matching set, sorting out the diagnostic results corresponding to each case, including information such as the core root cause of the fault, the scope of influence, and supplementary optimization directions, and comparing the feature vector of the current optimization effect to analyze the correlation between the diagnostic results of historical cases and the current optimization effect. For example, whether the root cause corresponding to the residual problem in the historical case is consistent with the possible root cause of the current residual problem, and whether the supplementary optimization strategy of the historical case is applicable to the current scenario. A preset similarity threshold is set based on the validity data of historical case matching to determine whether historical cases have reference value. The comprehensive similarity of cases in the historical pattern matching set is checked one by one. If there is a historical optimization case with a comprehensive similarity exceeding the preset threshold, it means that the current optimization effect is highly similar to that historical case, and the diagnostic result has direct reference significance. Therefore, the diagnostic result of the historical case is inherited as a secondary diagnostic result, including supplementary root cause analysis and optimization suggestions. If the comprehensive similarity of all cases in the historical pattern matching set does not reach the preset threshold, it means that the current optimization scenario or residual problem is special and there is no directly referable historical experience. At this time, based on the residual problem indication in the current optimization effect evaluation data, combined with the operation status data of the distributed order processing cluster and the previous diagnostic results, the core root cause of the residual problem is re-analyzed to clarify the scope and severity of the problem's impact and generate a targeted secondary diagnostic result.

[0102] In this embodiment of the invention, by extracting optimization success identifiers, effect level scores, and residual problem indicators from the optimization effect evaluation data to construct a feature vector of the current optimization effect, and by performing similarity retrieval and sorting filtering in the historical diagnostic result database based on this feature vector to form a historical pattern matching set, and by combining a preset similarity threshold to select to inherit the diagnostic results of historical cases with high matching degree or to regenerate based on residual problem indicators, the technical means of traditional secondary diagnosis effectively overcome the technical problems of insufficient reuse of historical operation and maintenance experience, repetitive and cumbersome diagnostic process, and low efficiency and insufficient accuracy in root cause location of residual problems. Thus, the technical effects of high efficiency in the secondary diagnosis process, accurate diagnostic results, and maximization of the value of historical operation and maintenance experience are achieved. This not only reduces the computing power consumption and time cost of repeated diagnosis, but also improves the reliability of root cause location of residual problems.

[0103] In a preferred embodiment of the present invention, step 6 above may include:

[0104] Step 6.1: Receive the secondary diagnostic results and extract the root cause confirmation status, optimization effect sustainability assessment, and potential risk warning information from the secondary diagnostic results to obtain secondary diagnostic result information. Specifically, this includes: First, receiving the output secondary diagnostic results. The results are targeted diagnostic information generated based on the optimization effect assessment data and historical diagnostic cases, comprehensively covering the problem resolution status, effect sustainability, and potential risk warnings. From the secondary diagnostic results, extract the root cause confirmation status to clarify whether the previously identified core root causes, such as memory leaks caused by JVM heap memory parameter mismatches, have been completely eliminated, or whether there are any new root causes not discovered in the initial diagnosis, and the impact of residual problems on... The root cause type has been identified; next, information on the sustainability of optimization effects is extracted, and the stability of key performance indicators such as node response time and database connection count after the optimization strategy is implemented is analyzed to determine whether the effect has diminished or rebounded, and to assess the long-term effectiveness of the optimization; finally, potential risk warning information is extracted, and combined with the dynamic operating characteristics of operation and maintenance objects such as distributed order processing clusters, potential hazards that may cause similar failures to recur, such as insufficient parameter adaptation due to fluctuations in instantaneous order volume peaks, or the probability of related failures, are identified. The extracted root cause, status, optimization effect sustainability assessment, and potential risk warning information are then organized to form complete secondary diagnostic results.

[0105] Step 6.2: Based on the extracted secondary diagnostic results and combined with the optimization execution data, calculate the deviation between the actual optimization effect indicators and the expected optimization goals. Analyze the temporal characteristics and impact dimensions of the deviation, and generate a deviation analysis report containing the deviation magnitude, type, and scope of influence. Specifically, this includes: based on the extracted secondary diagnostic results, clarifying the expected optimization goals of the optimization strategy. These goals are set in conjunction with the normal operation standards of the maintenance object, such as keeping node response time within a reasonable range and stabilizing the number of database connections within a preset threshold, and fault handling requirements. This includes the target value range of key performance indicators and the duration of the optimization effect. Retrieve the generated optimization execution data, which records the entire process information of strategy execution, including execution parameters, operation steps, node feedback, etc. Analyze the actual optimization effect indicators reflected in the optimization execution data, such as the stable value of the optimized node response time and the fluctuation range of the number of database connections. The utilization rate of M-heap memory, etc., is compared with the preset expected optimization targets one by one. The deviation value between the actual value and the target value of each indicator is calculated. A positive deviation value indicates that the expected target has not been achieved, while a negative deviation value or a deviation value close to zero indicates that the expected target has been achieved or exceeded. Further analysis is conducted on the temporal characteristics of the deviation, i.e., whether the deviation exists at the beginning of the optimization execution, gradually appears after the effect has been stable for a period of time, or occurs periodically with changes in the operating environment such as fluctuations in order volume. At the same time, the impact dimensions of the deviation are clarified, determining whether the deviation only affects some distributed nodes or global nodes, whether it only involves a single business module or multiple related modules, and whether it will be transmitted to other key performance indicators. Based on the combined impact dimensions of the temporal characteristics of the deviation value, a deviation analysis report is generated, which includes the deviation magnitude (the difference between the actual value and the target value), the deviation type (persistent deviation, periodic deviation, sudden deviation), the deviation impact range (the number of nodes affected), and the business modules, etc.

[0106] Step 6.3: Based on the deviation analysis report, identify the key factors causing the deviation, including improper configuration of optimization strategy parameters, dynamic changes in the execution environment, or insufficient root cause localization accuracy. Determine the type and direction of the optimization parameters that need to be adjusted. Specifically, this includes: performing in-depth analysis of the generated deviation analysis report, and combining the secondary diagnostic results and optimization execution data to investigate the key factors causing the deviation one by one. If the deviation is caused by unreasonable settings of core parameters in the optimization strategy, such as the initial and maximum values ​​of JVM heap memory being adjusted too little to meet the peak demand of instantaneous order volume, it is determined to be due to improper configuration of optimization strategy parameters. If the deviation is caused by dynamic changes in the operating environment of the operation and maintenance object, such as the order volume during the promotion peak period far exceeding the historical peak, the addition or removal of distributed nodes causing changes in resource distribution, and the optimization strategy is not adapted in time, it is determined to be due to dynamic changes in the execution environment. If the deviation is caused by the initial root cause location not fully covering all key factors, such as only focusing on JVM heap memory parameters while ignoring the linkage effect of database connection pool parameters, resulting in insufficient targeting of the optimization strategy, it is determined to be due to insufficient root cause location accuracy. After identifying the key factors, combined with the type and scope of the deviation, determine the type of optimization parameters that need to be adjusted. For example, for improper parameter configuration, the step size of resource adjustment threshold parameters needs to be adjusted; for dynamic changes in the execution environment, the strategy execution frequency and associated strategy triggering conditions need to be adjusted; for insufficient root cause location accuracy, the fault feature matching parameters need to be adjusted, etc. At the same time, the direction of parameter adjustment is determined according to the deviation magnitude and timing characteristics.

[0107] Step 6.4: Based on key deviation factors, dynamically adjust the core parameters in the optimization strategy, including resource adjustment thresholds, strategy execution frequency, parameter adjustment step size, and associated strategy triggering conditions, to generate an adjusted optimization parameter configuration set. Specifically, this includes: dynamically adjusting the parameter types and directions based on identified key factors; for resource adjustment thresholds, appropriately raising or lowering the threshold standard according to the deviation magnitude and the resource carrying capacity of the operation and maintenance object; for example, when JVM heap memory usage continuously exceeds the target range, increasing the maximum heap memory threshold to ensure resource allocation can meet peak business demands; and for strategy execution frequency, adjusting the strategy based on the timing characteristics of the deviation and the frequency of changes in the operating environment. For example, when order volume fluctuates frequently, the execution interval should be shortened to achieve more real-time adaptation and adjustment. Regarding parameter adjustment step size, the step size should be optimized based on the rate of change of deviation and the stability requirements of the maintained object to avoid excessive adjustment leading to system fluctuations or insufficient adjustment failing to quickly eliminate deviations. For associated strategy trigger conditions, more sensitive trigger thresholds should be set based on potential risk warning information. For example, when the number of database connections approaches the warning value, the connection pool parameter adjustment strategy should be triggered in advance to prevent failures. During the adjustment process, the linkage between various parameters should be fully considered to avoid conflicts between parameter adjustments, ensuring that the adjusted parameter configurations are mutually compatible and form a logically self-consistent set of optimized post-adjustment parameter configurations.

[0108] Step 6.5: Based on the adjusted optimized parameter configuration set, update the corresponding strategy parameters in the preset optimization strategy knowledge base, and feed the updated strategy parameters back to the optimization strategy matching stage, triggering a new round of diagnosis, optimization, and verification processes. This forms a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis. Specifically, this includes: associating the adjusted optimized parameter configuration set with the preset optimization strategy knowledge base, locating the original strategy parameters in the knowledge base corresponding to the fault type, such as memory leak fault or connection anomaly fault, replacing the original parameters with the adjusted parameters, and simultaneously recording the reasons for the parameter adjustment, i.e., the core conclusions of the deviation analysis report. Regarding the adjustment time and expected effects, the index information and association rules of the strategies in the knowledge base are updated to ensure that the optimized strategies can be accurately matched during subsequent searches. The updated strategy parameters are transmitted to the optimization strategy matching stage through the internal feedback mechanism of the system, serving as the basic data for a new round of strategy matching. When the operation and maintenance object experiences a similar failure again or the running status data reaches the strategy triggering condition, the optimal strategy will be quickly matched based on the updated parameters, and a new round of diagnosis, optimization and verification process will be initiated. Through the cyclical mechanism of parameter adjustment, knowledge base update, strategy matching feedback, a closed-loop adaptive optimization process is formed from diagnosis to optimization and from verification feedback to re-diagnosis.

[0109] In this embodiment of the invention, by employing techniques such as extracting root cause confirmation status, continuous evaluation of optimization effects, and potential risk warning information from secondary diagnostic results, combining optimization execution data to calculate the deviation between the actual optimization effect and the expected target, analyzing the temporal characteristics and impact dimensions of the deviation, identifying key deviation factors such as improper configuration of optimization strategy parameters and dynamic changes in the execution environment, dynamically adjusting core parameters such as resource adjustment thresholds and strategy execution frequency, updating the optimization strategy knowledge base, and feeding the adjusted parameters back to the strategy matching stage to trigger a new round of closed-loop process, this invention effectively overcomes the technical problems in traditional operation and maintenance optimization, such as the lack of scientific deviation analysis support for parameter adjustment, static and fixed optimization parameters that cannot adapt to dynamic operating environments, and the lack of a closed-loop feedback mechanism that makes it difficult to continuously iterate and improve optimization effects. This achieves the technical effects of precise optimization parameter adjustment, dynamic adaptation of optimization strategies, and continuous iteration of the optimization knowledge base, forming a complete closed-loop adaptive optimization process. This improves the adaptability of operation and maintenance software to complex dynamic operating environments and the stability and continuity of optimization effects, while reducing the recurrence rate of faults and the cost of manual intervention.

[0110] like Figure 2 As shown, embodiments of the present invention also provide an adaptive optimization system for operation and maintenance software based on intelligent diagnosis, including:

[0111] The acquisition module is used to collect real-time operational status data of the maintenance objects, organize the operational status data into a multi-dimensional data set, and construct a data feature distribution structure based on the multi-dimensional data set. In the data feature distribution structure, the baseline analysis range is defined, and according to the inherent characteristics of the data, the high-dimensional baseline analysis range is recursively divided into eight trigram nodes. The division level is dynamically adjusted according to the amount of data and the uniformity of data distribution within each trigram, resulting in multiple trigram-type feature analysis areas.

[0112] The diagnostic module is used to collect various operating status data into the corresponding limit feature analysis area, generate feature weights based on the feature deviation and data correlation of each area, and perform real-time analysis of the operating status data based on the preset fault feature library and the feature weights to obtain diagnostic results.

[0113] The optimization module is used to automatically match and execute corresponding optimization strategies based on the root cause localization information in the diagnostic results, and obtain optimization execution data.

[0114] The monitoring module is used to continuously monitor the changing trends of key performance indicators of the operation and maintenance objects based on the optimization execution data during and after the optimization strategy is executed, so as to obtain monitoring results; and to perform time series correlation analysis on the monitoring results and diagnostic results to obtain optimization effect evaluation data.

[0115] The matching module is used to perform pattern matching between the current optimization effect and the historical diagnostic result database based on the optimization effect evaluation data to obtain the secondary diagnostic result;

[0116] The adjustment module is used to perform deviation analysis based on the secondary diagnostic results and the optimized execution data, dynamically adjust the optimization parameters, and form a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis.

[0117] The adaptive optimization system according to embodiments of the present invention can correspond to performing the method described in the embodiments of the present invention, and the above and other operations and / or functions of each module of the adaptive optimization system are respectively for implementing Figure 1 The corresponding process of the method in the illustrated embodiment will not be described in detail here for the sake of brevity.

[0118] This application also provides a computing device. This computing device can utilize a server.

[0119] like Figure 3 As shown in the figure, this is a schematic diagram of a computing device provided in an embodiment of this application. The computing device 700 includes a bus 701, a processor 702, a communication interface 703, and a memory 704. The processor 702, the memory 704, and the communication interface 703 communicate with each other via the bus 701.

[0120] The 701 bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0121] The processor 702 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).

[0122] Communication interface 703 is used for external communication. Memory 704 may include volatile memory, such as random access memory (RAM). Memory 704 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD). Executable code is stored in memory 704, and processor 702 executes this executable code to perform the aforementioned adaptive optimization prediction method for operation and maintenance software based on intelligent diagnostics.

[0123] Specifically, in implementing the embodiments of the intelligent diagnostic-based adaptive optimization system for operation and maintenance software described above, and where each module or unit of the intelligent diagnostic-based adaptive optimization system for operation and maintenance software described above is implemented in software, the software or program code required to execute the functions of each module / unit in the intelligent diagnostic-based adaptive optimization system for operation and maintenance software described above can be partially or entirely stored in the memory 704. The processor 702 executes the program code corresponding to each unit stored in the memory 704 to execute the aforementioned intelligent diagnostic-based adaptive optimization prediction method for operation and maintenance software.

[0124] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct the computing device to execute the aforementioned adaptive optimization prediction method for operation and maintenance software based on intelligent diagnosis.

[0125] This application also provides a computer program product comprising one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application are generated.

[0126] The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another website, computer, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0127] When the computer program product is executed by a computer, the computer executes any of the aforementioned methods of the intelligent diagnostic-based adaptive optimization prediction method for operation and maintenance software. The computer program product can be a software installation package; when any of the aforementioned methods of the intelligent diagnostic-based adaptive optimization prediction method for operation and maintenance software needs to be used, the computer program product can be downloaded and executed on the computer.

[0128] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An adaptive optimization method for operation and maintenance software based on intelligent diagnosis, characterized in that, The method includes: Real-time collection of operational status data of maintenance objects is used to organize the operational status data into a multi-dimensional dataset. A data feature distribution structure is constructed based on the multi-dimensional dataset. The baseline analysis range is defined in the data feature distribution structure. According to the inherent characteristics of the data, the high-dimensional baseline analysis range is recursively divided into eight trigram nodes. The division level is dynamically adjusted according to the amount of data and the uniformity of data distribution within each trigram, resulting in multiple trigram-type feature analysis areas. The data of each operating status are collected into the corresponding limit feature analysis area, and feature weights are generated based on the feature deviation and data correlation of each area. Based on the preset fault feature library and combined with the feature weights, the operating status data is analyzed in real time to obtain diagnostic results. Based on the root cause localization information in the diagnostic results, the corresponding optimization strategy is automatically matched and executed to obtain optimization execution data; Based on the optimized execution data, the key performance indicators of the operation and maintenance objects are continuously monitored during and after the optimization strategy is executed to obtain monitoring results; the monitoring results and diagnostic results are then subjected to time series correlation analysis to obtain optimization effect evaluation data. Based on the optimization effect evaluation data, the current optimization effect is matched with the historical diagnostic result database to obtain the secondary diagnostic result; Based on the deviation analysis between the secondary diagnosis results and the optimized execution data, the optimization parameters are dynamically adjusted to form a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis.

2. The adaptive optimization method for operation and maintenance software based on intelligent diagnosis according to claim 1, characterized in that, Real-time collection of operational status data of objects under maintenance; organization of operational status data into multi-dimensional datasets; and construction of data feature distribution structure based on multi-dimensional datasets. The baseline analysis range is defined within the data feature distribution structure. Based on the inherent characteristics of the data, the high-dimensional baseline analysis range is recursively divided into eight quaternary nodes. The division hierarchy is dynamically adjusted according to the amount of data and the uniformity of data distribution within each quaternary node, resulting in multiple quaternary feature analysis regions, including: By collecting real-time operational status data of the objects under maintenance and organizing the operational status data into a multi-dimensional dataset, a data feature distribution structure is constructed based on the multi-dimensional dataset. In the data feature distribution structure, the benchmark analysis range is defined based on the overall distribution range and variance of the operational status data; Based on the benchmark analysis scope, according to the inherent characteristics of the data, the high-dimensional benchmark analysis scope is recursively divided into eight octant nodes, where each octant node represents an orthogonal subspace initialized based on data density and covariance. Based on the partitioned nodes, the partitioning level is dynamically adjusted according to whether the amount of data in each node exceeds a preset threshold and whether the data distribution uniformity is lower than a preset uniformity threshold. Through recursive partitioning, each node meets the requirements of data volume balance and distribution uniformity, so as to obtain multiple partitioned feature analysis areas.

3. The adaptive optimization method for operation and maintenance software based on intelligent diagnosis according to claim 2, characterized in that, The data of each operating status are collected into the corresponding trigram feature analysis area, and feature weights are generated based on the feature deviation and data correlation of each area; Based on a pre-defined fault feature library and combined with feature weights, real-time analysis of operational status data is performed to obtain diagnostic results, including: Based on the obtained multiple trigram-type feature analysis regions, the data of each operating state are collected into the corresponding trigram-type feature analysis region to obtain the collection results; Based on the aggregation results, feature weights representing the importance of each feature are obtained according to the feature deviation of the data within each quaternary feature analysis area and the data correlation between different analysis areas. Based on feature weights and combined with a pre-set fault feature library, the collected operating status data is subjected to weighted real-time analysis to obtain the results of weighted real-time analysis. Based on the results of weighted real-time analysis, diagnostic results of root cause localization information are obtained.

4. The adaptive optimization method for operation and maintenance software based on intelligent diagnosis according to claim 3, characterized in that, Based on the root cause localization information in the diagnostic results, the corresponding optimization strategy is automatically matched and executed to obtain optimization execution data, including: Based on the diagnostic results, the root cause localization information is extracted from the diagnostic results. The root cause localization information includes the fault type identifier, the scope of impact parameters, and the severity level. Based on the extracted root cause localization information, multi-dimensional matching processing is performed in the preset optimization strategy knowledge base. The strategy category is determined according to the fault type identifier, the applicable strategy range is filtered in combination with the impact range parameter, and the strategy execution priority is determined according to the severity level to obtain the matching optimization strategy set. Based on the optimized strategy set, strategy conflict detection and fusion processing are performed. When there are multiple optional strategies, a comprehensive evaluation is conducted based on strategy execution efficiency, resource consumption cost and historical success rate data, and the final strategy is selected from the matching optimized strategy set as the target execution strategy. Based on the target execution strategy, a corresponding strategy execution instruction set is generated and distributed to the corresponding operation and maintenance object execution node. The execution node is controlled to perform parameter adjustment, service process restart and computing resource allocation operations according to the content of the instruction set. Based on the distributed policy execution instruction set and execution process, policy execution status data is collected in real time and integrated into optimized execution data.

5. The adaptive optimization method for operation and maintenance software based on intelligent diagnosis according to claim 4, characterized in that, Based on the optimized execution data, the key performance indicators of the operation and maintenance objects are continuously monitored during and after the optimization strategy is executed in order to obtain monitoring results; Time-series correlation analysis was performed between monitoring results and diagnostic results to obtain optimization effect evaluation data, including: Based on the optimized execution data, the execution nodes of the operation and maintenance objects to be monitored and their corresponding monitoring time windows are determined. Within the monitoring time window, the key performance indicators of the execution nodes of the operation and maintenance objects are sampled in real time to obtain the time series data of the key performance indicators. Based on the time series data of key performance indicators, the magnitude of change, rate of change and duration of stable state of each key performance indicator before and after the execution of the optimization strategy are calculated to obtain the trend curve of key performance indicator change, which is used as the monitoring result. Based on the monitoring results and combined with the diagnostic results, the trend curves of key performance indicators in the monitoring results are aligned with the root cause localization information in the diagnostic results according to the timestamps to construct a time series correlation matrix. Based on the time series correlation matrix, we analyze the causal relationship strength, time delay response characteristics and recovery degree between the changing trends of key performance indicators and root cause localization information, and calculate the quantitative indicators of optimization effect. Based on the quantitative indicators of optimization effect, combined with the preset effect evaluation threshold, optimization effect evaluation data is obtained, which includes optimization success indicators, effect level scores, and residual problem indicators.

6. The adaptive optimization method for operation and maintenance software based on intelligent diagnosis according to claim 5, characterized in that, Based on the optimization effect evaluation data, the current optimization effect is matched with the historical diagnostic result database to obtain secondary diagnostic results, including: Receive optimization effect evaluation data, extract optimization success identifiers, effect level scores and residual problem indicators from the evaluation data, and use them as the current optimization effect feature vector; Based on the feature vector of the current optimization effect, retrieve the historical optimization case with the highest similarity to the feature vector of the current optimization effect from the preset historical diagnostic result database; Based on the retrieved historical optimization cases, the feature vector of the current optimization effect is calculated and the feature similarity between the current optimization effect and each historical optimization case is calculated. The top N most similar historical optimization cases are selected according to the similarity ranking to obtain the historical pattern matching set. Based on the historical pattern matching set, the correlation between the diagnostic results of each historical optimization case and the current optimization effect is analyzed. When there are historical optimization cases in the historical pattern matching set with similarity exceeding a preset threshold, the diagnostic results of the historical optimization cases are inherited as secondary diagnostic results. When there are no historical optimization cases that meet the similarity threshold, secondary diagnostic results are regenerated based on the residual problem indications in the current optimization effect evaluation data.

7. The adaptive optimization method for operation and maintenance software based on intelligent diagnosis according to claim 6, characterized in that, Based on the deviation analysis between the secondary diagnostic results and the optimized execution data, the optimization parameters are dynamically adjusted to form a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis, including: Receive the secondary diagnostic results, extract the root cause confirmation status, continuous evaluation of optimization effect, and potential risk warning information from the secondary diagnostic results, and obtain the secondary diagnostic result information; Based on the extracted secondary diagnostic results and combined with the optimization execution data, the deviation between the actual optimization effect index and the expected optimization target is calculated. The temporal characteristics and impact dimensions of the deviation are analyzed, and a deviation analysis report containing the deviation magnitude, deviation type and deviation impact range is generated. Based on the deviation analysis report, identify the key factors that cause the deviation, including improper configuration of optimization strategy parameters, dynamic changes in the execution environment, or insufficient root cause localization accuracy, and determine the type and direction of optimization parameters that need to be adjusted. Based on the key factors of deviation, the core parameters in the optimization strategy are dynamically adjusted, including resource adjustment threshold, strategy execution frequency, parameter adjustment step size and associated strategy triggering conditions, to generate an adjusted set of optimization parameter configurations. Based on the adjusted optimized parameter configuration set, the corresponding strategy parameters in the preset optimization strategy knowledge base are updated, and the updated strategy parameters are fed back to the optimization strategy matching stage, triggering a new round of diagnosis, optimization, and verification processes, forming a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis.

8. An adaptive optimization system for operation and maintenance software based on intelligent diagnosis, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to collect real-time operational status data of the maintenance objects, organize the operational status data into a multi-dimensional data set, and construct a data feature distribution structure based on the multi-dimensional data set. In the data feature distribution structure, the baseline analysis range is defined. Based on the inherent characteristics of the data, the high-dimensional baseline analysis range is recursively divided into eight quagmire nodes. The division level is dynamically adjusted according to the amount of data and the uniformity of data distribution in each quagmire, resulting in multiple quagmire-type feature analysis areas. The diagnostic module is used to collect data from various operating states into the corresponding trigram feature analysis areas and generate feature weights based on the feature deviation and data correlation of each area. Based on a pre-set fault feature library and combined with feature weights, the operating status data is analyzed in real time to obtain diagnostic results; The optimization module is used to automatically match and execute corresponding optimization strategies based on the root cause localization information in the diagnostic results, and obtain optimization execution data. The monitoring module is used to continuously monitor the changing trends of key performance indicators of the operation and maintenance objects based on the optimization execution data during and after the optimization strategy is executed, so as to obtain monitoring results; and to perform time series correlation analysis on the monitoring results and diagnostic results to obtain optimization effect evaluation data. The matching module is used to perform pattern matching between the current optimization effect and the historical diagnostic result database based on the optimization effect evaluation data to obtain the secondary diagnostic result; The adjustment module is used to perform deviation analysis based on the secondary diagnostic results and the optimized execution data, dynamically adjust the optimization parameters, and form a closed-loop adaptive optimization process from diagnosis to optimization, and then from verification feedback to re-diagnosis.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images