A method and system for constructing a membrane material database

Abstract

The present invention discloses a method and system for constructing a membrane material database, which relates to the technical field of database construction, and includes the following steps: in the operation process of the database management system, multi-dimensional parameter information generated by the real-time acquisition system is collected, the collected raw data is cleaned and preprocessed, and different types of data are normalized to ensure the consistency and validity of the data. The present invention collects multi-dimensional parameters of the database management system in real time and performs cleaning and abnormal analysis, and the system intelligently evaluates the operating status, generates an abnormal operation coefficient, and realizes early warning and fault prediction. Based on the comparison between the abnormal coefficient and the multi-level threshold, the system divides the risk into low, medium and high levels, and takes corresponding measures. This refined management mechanism effectively improves the stability of the database, prevents data loss, inconsistency and collapse, and improves the accuracy and efficiency of fault handling.

Description

Technical Field

[0001] The present invention relates to the technical field of database construction, and in particular to a method and system for constructing a membrane material database. Background Art

[0002] The construction method for membrane material database refers to building a systematic and structured database by collecting, organizing and storing various types of information related to membrane materials, so as to provide researchers and engineers with data support such as the performance, application, and preparation methods of membrane materials. First of all, the construction of the database requires clear data sources, including scientific literature, experimental data, patent information, and industry standards. Through data mining, manual input, or automated tool capture, information on the physical properties (such as pore size, permeability, selectivity), chemical composition, manufacturing process, and application scenarios of membrane materials is collected. Secondly, the standardized processing of data is crucial. The data needs to be classified, labeled, and structured for storage to ensure the accuracy, completeness, and retrievability of the data, so as to facilitate subsequent data query, analysis, and modeling. The construction process can also use database management systems (DBMS) and data analysis tools such as SQL and Python to achieve effective management and application of data.

[0003] The construction method of membrane material database requires not only accurate extraction of raw data, but also multi-dimensional classification and retrieval functions through reasonable database architecture design. For example, by setting different attribute labels for membrane materials (such as material type, purpose, physical properties, etc.), users can filter and compare according to specific needs to obtain the most relevant data. In addition, advanced database systems can also combine machine learning and artificial intelligence algorithms to provide data prediction and analysis functions to help researchers develop and optimize new materials. Continuous updating and maintenance of the database is also critical to ensure that the latest scientific research results and technological advances can be incorporated and applied to actual projects in a timely manner.

[0004] The prior art has the following deficiencies:

[0005] The database management system (DBMS) is the core of the normal operation of the membrane material database, responsible for ensuring the efficient and accurate execution of data operations, management, retrieval and maintenance functions. Once an abnormality occurs in the DBMS and is not detected in time, key functions such as data retrieval, storage, update, backup and recovery may be damaged, leading to data loss, inconsistency, and even system crash.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system for constructing a membrane material database. By collecting multi-dimensional parameter information of a database management system (DBMS) in real time and performing cleaning, preprocessing and abnormality analysis, the system can intelligently evaluate the operating status, generate an operating abnormality coefficient, and achieve early warning and fault prediction of high-risk abnormalities. Based on the comparative analysis of the operating abnormality coefficient and the multi-level threshold, the system divides the abnormalities into low, medium and high risks, and takes corresponding measures for different risk levels. This refined risk management mechanism effectively improves the stability of the database, prevents data loss, inconsistency and system crashes, and significantly improves the accuracy and efficiency of fault handling to solve the problems in the above-mentioned background technology.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a method for constructing a membrane material database, comprising the following steps:

[0009] During the operation of the database management system, the multi-dimensional parameter information generated by the system is collected in real time, the collected raw data is cleaned and preprocessed, and different types of data are standardized to ensure the consistency and validity of the data;

[0010] Perform preliminary anomaly detection on the preprocessed operating parameters, identify abnormal points in the operating status, and determine whether the operating parameters deviate from the normal operating range;

[0011] Input the parameter information processed by abnormal analysis into the pre-learned machine learning model, and judge the current operating status of the system according to the current parameter change trend;

[0012] Further comprehensive analysis is conducted on the database management system with existing abnormal operating states, and the abnormal operating states of the data management library are further divided. Based on the results of the division, different response measures are taken for different abnormal operating states.

[0013] Preferably, the multi-dimensional parameter information generated during the operation of the database management system includes connection pool saturation information and deadlock recovery information. The connection pool saturation information reflects the resource usage of the connection pool in the database management system. The deadlock recovery information refers to the relevant information on how to detect, process and recover when a deadlock occurs during the operation of concurrent transactions in the database system.

[0014] Preferably, after obtaining the connection pool saturation information and deadlock recovery information generated during the operation of the database management system, the connection pool saturation information and the deadlock recovery information are subjected to exception analysis and processing, and a connection pool saturation index and a deadlock recovery index are generated respectively, and the operation process of the database management system is intelligently judged through the connection pool saturation index and the deadlock recovery index.

[0015] Preferably, after obtaining the connection pool saturation index and deadlock recovery index generated during the operation of the database management system, the connection pool saturation index and the deadlock recovery index are input into a pre-learned machine learning model, and an operation anomaly coefficient is generated by the machine learning model, and the operation process of the database management system is judged by the operation anomaly coefficient.

[0016] Preferably, the operation abnormality coefficient generated when the database management system is running is often compared and analyzed with the preset operation abnormality coefficient reference threshold, and the comparison and analysis results are as follows:

[0017] If the operation abnormality coefficient is greater than or equal to the preset operation abnormality coefficient reference threshold, it is judged that the operation process of the database management system is in an abnormal operation state, and a high-risk abnormal operation signal is generated;

[0018] If the operation abnormality coefficient is less than a preset operation abnormality coefficient reference threshold, it is judged that the operation process of the database management system is in a stable operation state, and a low-risk abnormal operation signal is generated.

[0019] Preferably, when a high-risk abnormal operation signal is generated within a fixed detection window during operation of the database management system, the operation abnormality coefficient generated within the fixed detection window during operation of the database management system is continuously obtained to establish an analysis set, and the operation abnormality coefficient in the analysis set is compared and analyzed with the first gradient reference threshold, the second gradient reference threshold, and the operation abnormality coefficient reference, wherein the second gradient reference threshold is greater than the first gradient reference threshold, and the first gradient reference threshold is greater than the operation abnormality coefficient reference threshold, and the second gradient reference threshold, the first gradient reference threshold, and the abnormality coefficient reference threshold are compared and analyzed with the operation abnormality coefficient, and the operation abnormality coefficient that is less than the first gradient reference threshold and greater than or equal to the operation abnormality coefficient reference threshold is calibrated as , the operation abnormality coefficient that is less than the second gradient reference threshold and greater than or equal to the first gradient reference threshold is calibrated as , calibrate the operation abnormality coefficient greater than or equal to the second gradient reference threshold ;

[0020] Will , , Conduct comprehensive analysis to generate abnormal operation coefficients , based on the formula: , where , , They are , , The preset scaling factor of , , Both are greater than 0.

[0021] Preferably, the abnormal operation coefficient generated by further analyzing the operation process of the database management system is compared and analyzed with the preset first abnormal operation coefficient reference threshold and the second abnormal operation coefficient reference threshold to determine the risk level of the abnormal operation. The comparison and analysis results are as follows:

[0022] If the abnormal operation coefficient is less than the first abnormal operation coefficient reference threshold, the risk level of the abnormal operation is classified as a low-risk abnormal operation state;

[0023] If the abnormal operation coefficient is greater than or equal to the first abnormal operation coefficient reference threshold and less than the second abnormal operation coefficient reference threshold, the risk level of the abnormal operation is classified as a medium-risk abnormal operation state;

[0024] If the abnormal operation coefficient is greater than or equal to the second abnormal operation coefficient reference threshold, the risk level of the abnormal operation is classified as a high-risk abnormal operation state.

[0025] Preferably, the logic of performing abnormal analysis on the connection pool saturation information to generate the connection pool saturation index is as follows:

[0026] In the detection window, collect real-time parameters related to connection pool resource usage, including the number of active connections in the connection pool. , the maximum number of available connections in the connection pool , the number of requests waiting to obtain a connection and how long the connection is maintained ;

[0027] Based on the collected parameters, the connection pool utilization and connection pool waiting rate at each time point are defined. Taking into account the influence of all factors, an exponential decay model is used to give weights to different time points. The calculation expression is as follows: , , where Indicates the connection pool utilization, Indicates the connection pool waiting rate, represents the natural base, Represents the exponential decay factor in the connection pool utilization In the calculation expression, Used to control the connection hold time Connection pool utilization The impact of the connection pool wait rate In the calculation expression, To control the holding time Impact on connection pool wait rate;

[0028] Introducing dynamic weight factors , capturing the dynamic fluctuations of connection pool resources in different time periods, and used to adjust the impact of different time periods. The calculation expression is as follows: ;

[0029] The severity of connection pool saturation is quantified by the abnormality degree, which combines the connection pool utilization rate and the connection pool waiting rate. The calculation expression is as follows: , where Indicates abnormality, Indicates the adjustment coefficient, which is used to balance the impact of utilization rate and waiting rate;

[0030] Aggregate the abnormality values ​​at all time points and calculate the overall connection pool saturation index. The calculation expression is as follows: , where Indicates the connection pool saturation index. The total number of time points within the detection window.

[0031] Preferably, the logic of performing abnormal analysis on deadlock recovery information to generate a deadlock recovery index is as follows:

[0032] In the detection window, first collect multiple dimensions of information related to deadlock during the operation of the database management system, including the frequency of deadlock occurrence , Total number of transaction rollbacks caused by deadlock , The total waiting time of all transactions during deadlock recovery , the number of transactions delayed or failed due to deadlock and deadlock recovery delay ;

[0033] Deadlock recovery efficiency depends not only on the deadlock frequency , and the number of transactions delayed or failed due to deadlocks , The total waiting time of all transactions during deadlock recovery and lock recovery delay Therefore, first construct a deadlock impact factor, the calculation expression is as follows: , where Represents the deadlock impact factor;

[0034] To further quantify the difficulty and complexity of deadlock recovery, a deadlock complexity factor is defined. Combining the number of transactions involved in the deadlock and the number of resource types involved, the calculation expression is as follows: , Indicates the number of transactions competing for resources in a deadlock event, representing the complexity of the deadlock. Indicates the number of resource types involved in the deadlock;

[0035] Finally, the deadlock impact factor and deadlock complexity factor Combined, the deadlock recovery index is generated, and the calculation expression is as follows: , where Indicates the deadlock recovery index.

[0036] A system for building a membrane material database, including a data acquisition and preprocessing module, an anomaly detection and identification module, a machine learning prediction and state assessment module, and an anomaly classification and response strategy module;

[0037] The data collection and preprocessing module collects multi-dimensional parameter information generated by the system in real time during the operation of the database management system, cleans and preprocesses the collected raw data, and standardizes different types of data to ensure data consistency and validity;

[0038] The anomaly detection and identification module performs preliminary anomaly detection on the preprocessed operating parameters, identifies abnormal points in the operating state, and determines whether the operating parameters deviate from the normal operating range;

[0039] The machine learning prediction and status assessment module inputs the parameter information after abnormal analysis into the pre-learned machine learning model and determines the current operating status of the system based on the current parameter change trend;

[0040] The abnormal classification and response strategy module further comprehensively analyzes the database management system with existing abnormal operating states, further divides the abnormal operating states of the data management library, and based on the results of the division, takes different response measures for different abnormal operating states.

[0041] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0042] The present invention can effectively improve the operational stability of the database by collecting multi-dimensional parameter information during the operation of the database management system (DBMS) in real time, and cleaning, preprocessing and abnormal analysis of these data. In particular, through the monitoring and analysis of connection pool saturation information and deadlock recovery information, the system can accurately identify abnormal points in operation. Then, the operation status of the database is intelligently evaluated through a machine learning model, and an operation abnormality coefficient is further generated and compared with a preset threshold. This scheme can generate a high-risk abnormal signal when an abnormality occurs, and judge the specific risk level of the system through the accurate classification of the abnormal operation coefficient, so as to achieve early warning and fault prediction of system abnormalities. This mechanism can ensure that problems can be prevented or responded to in advance before they occur, thereby significantly reducing the risk of data loss, inconsistency and system crash, and effectively ensuring the stable operation of the database.

[0043] The present invention realizes hierarchical management of database operation anomalies by comparing and analyzing the operation anomaly coefficient with multi-level thresholds. By classifying anomalies into low risk, medium risk and high risk, the system can take corresponding countermeasures according to different risk levels. For low-risk anomalies, the system can optimize operation through automatic adjustment; for medium-risk anomalies, the system can take preventive repair measures; and for high-risk anomalies, the system can trigger emergency response strategies in time to avoid system crashes. Through this refined risk management, the database management system can flexibly deploy resources according to the severity of the anomaly, and take effective solutions according to different risk states, greatly improving the accuracy and efficiency of system fault handling. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0045] Figure 1 The present invention is a method flow chart of a method for constructing a membrane material database.

[0046] Figure 2 The module diagram of a system for constructing a membrane material database according to the present invention is shown in FIG. DETAILED DESCRIPTION

[0047] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0048] The present invention provides Figure 1 A method for constructing a membrane material database is shown, comprising the following steps:

[0049] During the operation of the database management system, the multi-dimensional parameter information generated by the system is collected in real time, the collected raw data is cleaned and preprocessed, and different types of data are standardized to ensure the consistency and validity of the data;

[0050] The multi-dimensional parameter information generated during the operation of the database management system includes connection pool saturation information and deadlock recovery information. The connection pool saturation information reflects the resource usage of the connection pool in the database management system. The deadlock recovery information refers to the relevant information on how to detect, process and recover when a deadlock occurs during the operation of concurrent transactions in the database system.

[0051] When processing the collected raw data, data cleaning and preprocessing operations are first required, which is a key step to ensure the accuracy of subsequent data analysis and modeling. Data cleaning includes removing outliers, filling missing data, correcting unreasonable numerical ranges, etc., to eliminate noise and errors in the data. At the same time, different types of data need to be standardized to ensure that data of different dimensions have unified dimensions and representation methods. This step not only improves the consistency of the data, but also enhances the comparability and operability of the data, avoiding deviations and errors caused by differences in data distribution or inconsistent formats, thereby laying a solid foundation for subsequent analysis and model training, and ensuring the validity and reliability of the data in the entire system.

[0052] Perform preliminary anomaly detection on the preprocessed operating parameters, identify abnormal points in the operating status, and determine whether the operating parameters deviate from the normal operating range;

[0053] After obtaining the connection pool saturation information and deadlock recovery information generated during the operation of the database management system, the connection pool saturation information and the deadlock recovery information are analyzed and processed abnormally, and a connection pool saturation index and a deadlock recovery index are generated respectively. The operation process of the database management system is intelligently judged through the connection pool saturation index and the deadlock recovery index;

[0054] During the operation of the database management system, changes in connection pool saturation information often indicate that the use of connection resources is close to saturation. When the connection pool resources are exhausted, new database requests cannot allocate connections, which will directly lead to abnormal operation of the database management system. Connection pool saturation will have a profound impact on the key functions of the database. First, data retrieval and storage operations will be blocked or delayed, query requests cannot be responded to in time, and data write operations may fail or time out. As resource consumption continues to increase, the reliability of data updates and transaction submissions will also decrease, and some transactions may not be successfully submitted, resulting in data inconsistency or loss. Since the resources in the connection pool cannot be effectively released, the system will further deteriorate under high load, which will hinder regular backup operations and make it impossible to effectively restore data in the event of a failure. Long-term connection pool saturation will also cause the performance of the entire system to collapse, and may eventually cause the system to crash. Therefore, abnormal changes in connection pool saturation information are an early warning of database performance degradation. If not handled in a timely manner, it will seriously affect the availability, data integrity and stability of the system, and may even cause the entire system to crash.

[0055] The logic for generating the connection pool saturation index by performing an exception analysis on the connection pool saturation information is as follows:

[0056] In the detection window, collect real-time parameters related to connection pool resource usage, including the number of active connections in the connection pool. , the maximum number of available connections in the connection pool , the number of requests waiting to obtain a connection and how long the connection is maintained ;

[0057] Based on the collected parameters, the connection pool utilization and connection pool waiting rate at each time point are defined. Taking into account the influence of all factors, an exponential decay model is used to give weights to different time points. The calculation expression is as follows: , , where Indicates the connection pool utilization, Indicates the connection pool waiting rate, represents the natural base, Represents the exponential decay factor in the connection pool utilization In the calculation expression, Used to control the connection hold time Connection pool utilization The impact of the connection pool wait rate In the calculation expression, To control the holding time Impact on connection pool wait rate;

[0058] It should be noted that:

[0059] The Exponential Decay Model is a mathematical model commonly used to weight time series data, in which older data points are given smaller weights over time, while newer data points are given larger weights. The basic feature of this model is that the decay rate of the weight is exponentially related to time, that is, the weight decreases rapidly in the form of an exponential function over time.

[0060] The characteristic of the exponential decay model is that the weight of recent data points is larger, and as time goes by, the weight decreases rapidly in an exponential form. This model is widely used in many scenarios, such as data monitoring, predictive modeling, signal processing, etc., to weaken the impact of historical data and enhance the sensitivity to recent data.

[0061] In database management systems, the exponential decay model can be used to weight time series data (such as connection pool utilization, waiting rate, etc.) so that different weights are given to data at different time points when calculating system performance indicators. Specifically, the role of the exponential decay model here is as follows:

[0062] 1. Reflect dynamic changes;

[0063] In database systems, the utilization and waiting rate of the connection pool change dynamically. If only simple average calculation or linear weight calculation is used, it is likely that the load status of the system at different times cannot be accurately reflected. By introducing the exponential decay model, the system can give higher weight to the data at more recent times and quickly respond to the current system load changes. This allows the utilization and waiting rate of the connection pool to more accurately reflect the current pressure status.

[0064] 2. Reduce the impact of historical data;

[0065] When a database system runs for a long time, earlier data points (such as connection pool utilization or waiting rate) may no longer reflect the current system status. If these historical data are used directly for calculations, the system will not be able to capture new load conditions in a timely manner. The exponential decay model performs exponential decay on older data points, causing their weights to decrease rapidly. This significantly weakens the impact of historical data when evaluating the system status, while the most recent data is strengthened, allowing the system to more sensitively capture current load changes.

[0066] 3. Avoid interference from short-term peaks;

[0067] Occasionally, there may be short-term connection peaks or fluctuations in the database system. The exponential decay model can prevent short-term abnormal peaks from excessively affecting the overall load judgment in this case. For example, the connection pool utilization rate may surge at a certain point in time, but the situation quickly returns to normal. The exponential decay model quickly decays historical data so that such instantaneous peaks will not have too much impact on system indicators for a long time, thereby avoiding false alarms.

[0068] 4. Improve the real-time performance of the system;

[0069] When monitoring connection pool utilization or waiting rate, real-time performance is a key requirement. The exponential decay model allows the system to pay more attention to recent changes rather than historical accumulated data, so that it can quickly respond to changes in connection pool pressure and make more timely decisions. For example, when the connection pool is about to be saturated, the exponential decay model allows the system to be sensitive to the rapid increase in recent loads, so that measures can be taken in advance to avoid system overload.

[0070] Introducing dynamic weight factors , capturing the dynamic fluctuations of connection pool resources in different time periods, and used to adjust the impact of different time periods. The calculation expression is as follows: ;

[0071] It should be noted that:

[0072] A dynamic weight factor is a coefficient that dynamically adjusts the weight according to the system state or time. It is used to give different influences to different factors or data points during the calculation process. This weight will be continuously adjusted as the system state, time or other specific conditions change to ensure that the model or system can flexibly respond to these changes and make reasonable decisions. Dynamic weight factors are often used to capture important time-varying characteristics in the system and enhance the adaptability of the model.

[0073] The role of dynamic weight factors;

[0074] Capturing system changes: The main function of the dynamic weight factor is to dynamically adjust the impact of different data points on the results based on the current system status. For example, in the connection pool utilization monitoring of a database management system, if the system load is high, the dynamic weight factor can amplify the impact of the recent number of connections, making the system more sensitive to pressure changes.

[0075] Flexible adaptation to time-varying data: Dynamic weight factors can adapt to the ever-changing data environment, allowing the model to process data in different time periods differently. Compared with static weights, dynamic weights can make the model more intelligent. For example, during peak hours, the dynamic weight factor will increase to strengthen the processing of data under the current high load, while during low-load periods, the weight factor may decrease to avoid overreaction to short-term fluctuations.

[0076] Enhance the real-time and responsiveness of the system: Dynamic weighting factors allow the system to respond quickly based on real-time monitoring data. It can highlight important moments or factors, making the system more flexible to respond to instantaneous changes or abnormal conditions. For example, when the system detects a rapid increase in pressure at a certain moment, the dynamic weighting factor will increase, allowing the system to respond to this change in a timely manner.

[0077] Balancing the impact of historical data and current data: In complex systems, it is necessary to consider historical data while adapting to current changes. Dynamic weighting factors can balance the two by adjusting weights. For example, historical data can help the system identify long-term trends, while current data can capture short-term fluctuations. Dynamic weighting factors can increase or decrease the weight of historical data at different times to find the optimal balance.

[0078] The severity of connection pool saturation is quantified by the abnormality degree, which combines the connection pool utilization rate and the connection pool waiting rate. The calculation expression is as follows: , where Indicates abnormality, Indicates the adjustment coefficient, which is used to balance the impact of utilization rate and waiting rate;

[0079] It should be noted that:

[0080] Abnormality is an indicator that measures the degree to which the system or data status deviates from normal operating conditions. It is usually used to monitor and evaluate whether the system has abnormal conditions. Abnormality is usually calculated based on the comprehensive calculation of certain key performance parameters (such as connection pool utilization, waiting rate, response time, etc.) and generated through formulas or models. It can quantify the abnormal conditions of the system at a certain point in time, thereby helping operation and maintenance personnel or automation systems to promptly discover potential problems and avoid further deterioration of faults.

[0081] The core function of abnormality is to provide a comprehensive indicator to measure whether the current state of the system exceeds the set normal range. In the database management system, the abnormality can be calculated by weighted calculation based on factors such as the usage of the connection pool, the number of active connections, the number of request waits, and resource utilization. Generally, the abnormality will increase with the increase of system load, the delay of response time, the intensification of resource contention and other problems, reflecting the deterioration of the system operating environment. Conversely, a decrease in abnormality indicates that the system is in a healthy state.

[0082] Aggregate the abnormality values ​​at all time points and calculate the overall connection pool saturation index. The calculation expression is as follows: , where Indicates the connection pool saturation index. The total number of time points within the detection window;

[0083] In the detection window, the connection pool saturation index calculation expression shows that the larger the connection pool saturation index performance value generated after abnormal analysis of the connection pool saturation information, the greater the connection resource pressure or abnormal situation the database management system (DBMS) is facing during operation. As the connection pool saturation index increases, the performance of key functions in the system such as data retrieval, storage, update, backup and recovery operations will be significantly affected. Insufficient connection resources will lead to problems such as query response time delays, transaction submission failures, data writing obstructions, and further may cause serious consequences such as data loss and data inconsistency. In extreme cases, long-term connection pool saturation may even cause the entire system to crash. Therefore, the larger the index value, the higher the risk faced by the system and the greater the probability of crash or serious failure. On the contrary, if the connection pool saturation index performance value is smaller, it means that the database management system is running smoothly, resource allocation is reasonable, the system is in a healthy state as a whole, and all key functions can operate normally.

[0084] Changes in lock recovery information during the operation of a database management system, especially when the deadlock frequency increases or the recovery efficiency decreases, may cause abnormal system operation, thereby affecting key functions such as data retrieval, storage, update, backup and recovery. When a deadlock occurs, multiple transactions wait for resources, causing the system to be unable to execute these transactions normally. If deadlocks occur frequently or the system deadlock recovery mechanism is not efficient enough, transaction delays will increase, which may cause some transactions to fail to commit normally or be forced to roll back. This will not only affect the response speed of data retrieval, but may also cause data storage and update operations to be interrupted, resulting in inconsistent or lost data. In addition, improper deadlock handling may affect the database backup and recovery process, hindering the system from performing effective fault recovery at critical moments. As the deadlock frequency increases, long-term resource locking will lead to system resource exhaustion, which may cause the overall system performance to decline or even crash. Therefore, abnormal changes in deadlock recovery information are directly related to the stability and data integrity of the system, and have a profound negative impact on database operation.

[0085] The logic of performing exception analysis on deadlock recovery information to generate deadlock recovery index is as follows;

[0086] In the detection window, first collect multiple dimensions of information related to deadlock during the operation of the database management system, including the frequency of deadlock occurrence , Total number of transaction rollbacks caused by deadlock , The total waiting time of all transactions during deadlock recovery , the number of transactions delayed or failed due to deadlock and deadlock recovery delay ;

[0087] Deadlock recovery efficiency depends not only on the deadlock frequency , and the number of transactions delayed or failed due to deadlocks , The total waiting time of all transactions during deadlock recovery and lock recovery delay Therefore, first construct a deadlock impact factor, the calculation expression is as follows: , where Represents the deadlock impact factor;

[0088] It should be noted that the deadlock impact factor is a key indicator used to quantify the impact of deadlock on the overall operation of the database system. It combines parameters from multiple dimensions to measure the impact of deadlock on system resources, transaction processing, and performance. When a deadlock occurs, multiple transactions lock each other, causing them to be unable to continue executing, thereby blocking the operation of other normal transactions. The deadlock impact factor takes into account the sum of the waiting time of all transactions during deadlock recovery, the total number of transaction rollbacks caused by deadlock, the deadlock recovery delay, and the number of transactions delayed or failed due to deadlock. Specifically, the sum of the waiting time of all transactions during deadlock recovery is Represents the degradation of system performance caused by deadlock; the total number of transaction rollbacks caused by deadlock It reflects the number of transactions that cannot be completed normally when the system is processing deadlock. The more rollbacks, the greater the impact on the overall operation of the system. Deadlock recovery delay It describes the time it takes for the system to identify and resolve deadlocks. The longer the delay, the lower the efficiency of the system in deadlock processing, and the more serious the resource occupation and transaction delays. The total waiting time of all transactions during deadlock recovery It reflects the impact of deadlock on other concurrent transactions. The higher the value, the greater the damage to the overall system performance. By combining these factors into a formula, the deadlock impact factor can provide system administrators with a clear quantitative indicator to evaluate the impact of deadlock on the operational health and efficiency of the database system, so as to take optimization measures to reduce the negative effects of deadlock.

[0089] To further quantify the difficulty and complexity of deadlock recovery, a deadlock complexity factor is defined. The calculation expression is as follows, combining the number of transactions involved in the deadlock and the number of resource types involved (such as tables, indexes, row-level locks, etc.): , Indicates the number of transactions competing for resources in a deadlock event, representing the complexity of the deadlock. Indicates the number of resource types involved in the deadlock;

[0090] It should be noted that:

[0091] In the calculation of the deadlock recovery index, Represents the number of transactions competing for resources in a deadlock event, which reflects the complexity of the deadlock. Specifically, It refers to the situation where multiple transactions involved in the deadlock are mutually dependent or compete for the same resources, resulting in mutual blocking. The more transactions there are, the more complex the deadlock is, and the more transaction conflicts the system needs to deal with, which increases the difficulty of recovering from the deadlock.

[0092] The following are Detailed explanation:

[0093] 1. Transactions in deadlock are interdependent;

[0094] When multiple transactions access the same resource (such as database tables, indexes, row locks, etc.) at the same time, if there is mutual dependence between the transactions and there is no reasonable resource release mechanism, a deadlock may occur. For example, transaction A occupies resource X and waits for resource Y, while transaction B occupies resource Y and waits for resource X. In this case, the two are waiting for each other, forming a deadlock. As the number of participating transactions increases, the deadlock chain becomes more complex, and the system needs to perform more operations to recover these transactions.

[0095] 2. The relationship between the number of transactions and recovery complexity;

[0096] This directly affects the complexity of the system's detection and handling of deadlocks. When a deadlock occurs, the system usually needs to roll back some transactions to break the deadlock. If the number of transactions involved in the deadlock is small, the system only needs to roll back a few of them to resolve the deadlock. When the deadlock rate is high, the system must analyze the dependencies between multiple transactions and select an appropriate rollback strategy to minimize the impact on other transactions. The more transactions there are, the longer the deadlock chain is, and the higher the cost of resolving the deadlock is.

[0097] 3. The amplification effect of transaction conflicts;

[0098] along with As the number of transactions increases, competition between transactions will cause more resources to be locked, which may affect other transactions that are not directly involved in the deadlock. For example, in a high-concurrency environment, a deadlock may indirectly block many other transactions, causing the overall performance of the system to degrade. When the system recovers these transactions, it needs to consume more computing resources and time.

[0099] 4. System load and resource usage;

[0100] The more transactions involved in a deadlock, the more system resources (such as locks, CPU, and memory) will be occupied until the deadlock is detected and resolved. It not only increases the complexity of deadlock recovery, but also has a negative impact on the effective use of system resources. If the deadlock lasts for a long time, the entire database system may experience performance bottlenecks due to resource exhaustion or response delays, or even cause the system to crash.

[0101] 5. Rollback and data consistency;

[0102] along with As the number of transactions increases, the number of transactions that the system may need to roll back will also increase. Rollbacks affect the transaction consistency of the system, especially when multiple transactions are dependent. In order to maintain data consistency and integrity, the system must adopt more complex strategies when recovering these transactions, which further increases the complexity of deadlock recovery.

[0103] It is a key parameter to measure the complexity of deadlocks, representing the number of transactions competing for resources. The more transactions there are, the more difficult it is for the system to recover from deadlocks, because the system needs to analyze more dependencies, roll back more transactions, and face the risk of higher resource usage and wider conflict scope. Therefore, It directly determines the complexity and time-consuming of the deadlock recovery process.

[0104] Finally, the deadlock impact factor and deadlock complexity factor Combined, the deadlock recovery index is generated, and the calculation expression is as follows: , where Indicates the deadlock recovery index.

[0105] From the deadlock recovery index calculation expression, it can be seen that the larger the deadlock recovery index performance value generated by abnormal analysis of deadlock recovery information, the more frequent deadlock problems and low recovery efficiency occur when the system processes concurrent transactions. In this case, system resources (such as locks, CPU, and memory) are occupied for a long time, and the interdependence between transactions leads to frequent blocking and rollback, which directly affects the core functions of the database, such as data retrieval, storage, update, backup, and recovery. When deadlocks occur frequently and cannot be processed in a timely and efficient manner, delays and failures in data operations will increase, which may lead to problems such as data loss and data inconsistency. If the system deadlock recovery index continues to rise, the overall system load will become uncontrollable, resources will be exhausted or locked for a long time, and the database system may eventually crash. This will not only affect the operational stability of the database, but may also bring serious business interruption risks to applications that rely on the system. On the contrary, if the deadlock recovery index performance value is smaller, it means that the system has a low deadlock incidence rate during concurrent processing, and the system is efficient in deadlock detection and recovery, and can quickly resolve resource conflicts. At this point, the database management system is in a stable operating state, all key functions can run efficiently, transactions are executed smoothly, data operations are not significantly disturbed, and system performance is maintained at a normal level. Therefore, the size of the deadlock recovery index directly reflects the health of the system. The smaller the index, the more stable the system operation and the lower the risk of damage to key functions.

[0106] Input the parameter information processed by abnormal analysis into the pre-learned machine learning model, and judge the current operating status of the system according to the current parameter change trend;

[0107] Get the connection pool saturation index generated during the operation of the database management system and deadlock recovery index After that, the connection pool saturation index and deadlock recovery index Input into the pre-learned machine learning model, and generate the operation abnormality coefficient through the machine learning model , by running the anomaly coefficient Judge the operation process of database management system;

[0108] The machine learning model is not limited here and can achieve the connection pool saturation index and deadlock recovery index Conduct comprehensive analysis to generate operation abnormality coefficients In order to realize the technical solution of the present invention, the present invention proposes a specific implementation scheme:

[0109] Operation abnormality coefficient The calculation formula is generated as follows: , where , They are connection pool saturation index and deadlock recovery index The preset scaling factor of , Both are greater than 0.

[0110] It can be seen from the calculation expression of the operation abnormality coefficient that the larger the connection pool saturation index performance value generated during the operation of the database management system, and the larger the deadlock recovery index performance value generated during the operation of the database management system, the larger the operation abnormality coefficient performance value, indicating that anomalies occur during the operation of the database management system, resulting in damage to key functions such as data retrieval, storage, update, backup and recovery, and further causing data loss, inconsistency, and even The greater the probability of system crashes, conversely, it means that the operation process of the database management system is in a stable state without abnormal conditions.

[0111] The operation abnormality coefficient generated during the operation of the database management system is often compared and analyzed with the preset operation abnormality coefficient reference threshold. The comparison and analysis results are as follows:

[0112] If the operation abnormality coefficient is greater than or equal to the preset operation abnormality coefficient reference threshold, it is judged that the operation process of the database management system is in an abnormal operation state, and a high-risk abnormal operation signal is generated;

[0113] If the operation abnormality coefficient is less than a preset operation abnormality coefficient reference threshold, it is judged that the operation process of the database management system is in a stable operation state, and a low-risk abnormal operation signal is generated.

[0114] Further comprehensive analysis is conducted on the database management system with existing abnormal operating states, and the abnormal operating states of the data management library are further divided. Based on the results of the division, different response measures are taken for different abnormal operating states.

[0115] When a high-risk abnormal operation signal is generated within a fixed detection window during operation of the database management system, the operation abnormality coefficient generated within the fixed detection window during operation of the database management system is continuously obtained to establish an analysis set, and the operation abnormality coefficient in the analysis set is compared and analyzed with the first gradient reference threshold, the second gradient reference threshold, and the operation abnormality coefficient reference, wherein the second gradient reference threshold is greater than the first gradient reference threshold, and the first gradient reference threshold is greater than the operation abnormality coefficient reference threshold, and the second gradient reference threshold, the first gradient reference threshold, and the abnormality coefficient reference threshold are compared and analyzed with the operation abnormality coefficient, and the operation abnormality coefficient that is less than the first gradient reference threshold and greater than or equal to the operation abnormality coefficient reference threshold is calibrated as , the operation abnormality coefficient that is less than the second ladder reference threshold and greater than or equal to the first ladder reference threshold is calibrated as , calibrate the operation abnormality coefficient greater than or equal to the second ladder reference threshold ;

[0116] Will , , Conduct comprehensive analysis to generate abnormal operation coefficients , based on the formula: , where , , They are , , The preset scaling factor of , , Both are greater than 0.

[0117] It can be seen from the calculation expression of the abnormal operation coefficient that the larger the value of the abnormal operation coefficient is, the more serious the abnormality that occurs during the operation of the database management system is, and vice versa, the smaller the value is, the less serious the abnormality that occurs during the operation of the database management system is.

[0118] The abnormal operation coefficient generated by further analyzing the operation process of the database management system is compared and analyzed with the preset first abnormal operation coefficient reference threshold and the second abnormal operation coefficient reference threshold to determine the risk level of abnormal operation. The comparison and analysis results are as follows:

[0119] If the abnormal operation coefficient is less than the first abnormal operation coefficient reference threshold, the risk level of the abnormal operation is classified as a low-risk abnormal operation state;

[0120] If the abnormal operation coefficient is greater than or equal to the first abnormal operation coefficient reference threshold and less than the second abnormal operation coefficient reference threshold, the risk level of the abnormal operation is classified as a medium-risk abnormal operation state;

[0121] If the abnormal operation coefficient is greater than or equal to the second abnormal operation coefficient reference threshold, the risk level of the abnormal operation is classified as a high-risk abnormal operation state.

[0122] For low-risk abnormal operation states, although some small abnormalities occur in the operation of the database management system, these abnormalities have little impact on the overall performance and data operations of the system and can usually be handled through monitoring and optimization. The response measures for this type of risk should focus on regular monitoring and small-scale adjustments. Specific measures include: strengthening real-time monitoring of system performance indicators to ensure that low-risk abnormalities do not deteriorate further; optimizing database resource allocation, adjusting connection pool size or caching strategy, and reducing resource contention; at the same time, increasing checks on related logs to ensure that there are no hidden potential problems. The overall goal is to prevent small problems from developing into major failures, maintain system stability, and avoid the accumulation of low-risk abnormalities.

[0123] The medium-risk abnormal operation status indicates that part of the database operation has been greatly affected, which may cause performance degradation or certain delays in certain functions. In this case, active intervention measures must be taken to prevent the risk from escalating. Specific measures may include: optimizing query execution plans, optimizing queries that run for a long time or occupy a lot of resources, and reducing the occupation of system resources; reallocating system resources, such as by expanding server capacity or increasing cache space to alleviate resource bottlenecks; at the same time, the transaction management mechanism can be reviewed and adjusted to ensure that there are not too many conflicts between concurrent transactions. In addition, temporary database load analysis is also required to check the root cause of the risk anomaly and repair it in time to ensure that the anomaly does not further evolve into a high-risk state.

[0124] A high-risk abnormal operation state means that the database management system is in a serious unstable state, which may cause system downtime, data loss, or critical functions to fail to work properly. At this time, emergency recovery and comprehensive review measures must be taken. System backup should be performed immediately to ensure the safety of current data, and then high-load operations should be suspended or concurrent requests should be temporarily limited to prevent the system from crashing due to resource exhaustion. If possible, failover should be performed to switch the database to a redundant server or backup server to ensure the continuous availability of the system. At the same time, emergency troubleshooting needs to be started immediately to analyze the specific causes of high-risk anomalies, such as deadlock problems, resource leaks, or hardware failures, and to take repair measures quickly. In a high-risk state, the focus is on preventing data corruption, business interruption, and mitigating losses by restoring critical functions.

[0125] The present invention provides Figure 2 A system for building a membrane material database is shown, comprising a data acquisition and preprocessing module, an anomaly detection and identification module, a machine learning prediction and state assessment module, and an anomaly classification and response strategy module;

[0126] The data collection and preprocessing module collects multi-dimensional parameter information generated by the system in real time during the operation of the database management system, cleans and preprocesses the collected raw data, and standardizes different types of data to ensure data consistency and validity;

[0127] The anomaly detection and identification module performs preliminary anomaly detection on the preprocessed operating parameters, identifies abnormal points in the operating state, and determines whether the operating parameters deviate from the normal operating range;

[0128] The machine learning prediction and status assessment module inputs the parameter information after abnormal analysis into the pre-learned machine learning model and determines the current operating status of the system based on the current parameter change trend;

[0129] The abnormal classification and response strategy module further comprehensively analyzes the database management system with existing abnormal operating states, further divides the abnormal operating states of the data management library, and based on the results of the division, takes different response measures for different abnormal operating states.

[0130] A method for constructing a membrane material database provided in an embodiment of the present invention is implemented by the above-mentioned system for constructing a membrane material database. The specific method and process of a system for constructing a membrane material database are detailed in the embodiment of the above-mentioned method for constructing a membrane material database, which will not be repeated here.

[0131] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for constructing a membrane material database, characterized in that: The following steps are involved: During the operation of the database management system, the multi-dimensional parameter information generated by the system is collected in real time, the collected raw data is cleaned and preprocessed, and different types of data are standardized to ensure the consistency and validity of the data; Perform preliminary anomaly detection on the preprocessed operating parameters, identify abnormal points in the operating status, and determine whether the operating parameters deviate from the normal operating range; Input the parameter information processed by abnormal analysis into the pre-learned machine learning model, and judge the current operating status of the system according to the current parameter change trend; Further comprehensive analysis is conducted on the database management system with abnormal operation status, and the abnormal operation status of the database management library is further divided. Based on the division results, different response measures are taken for different abnormal operation statuses; The multi-dimensional parameter information generated during the operation of the database management system includes connection pool saturation information and deadlock recovery information. The connection pool saturation information reflects the resource usage of the connection pool in the database management system. The deadlock recovery information refers to how the database system detects, handles and recovers deadlocks when concurrent transactions are running. After obtaining the connection pool saturation information and deadlock recovery information generated during the operation of the database management system, the connection pool saturation information and the deadlock recovery information are analyzed and processed abnormally, and a connection pool saturation index and a deadlock recovery index are generated respectively. The operation process of the database management system is intelligently judged through the connection pool saturation index and the deadlock recovery index; After obtaining the connection pool saturation index and deadlock recovery index generated during the operation of the database management system, the connection pool saturation index and the deadlock recovery index are input into a pre-learned machine learning model, and an operation anomaly coefficient is generated by the machine learning model. The operation process of the database management system is judged by the operation anomaly coefficient.

2. A method for constructing a membrane material database according to claim 1, characterized in that: The operation abnormality coefficient generated during the operation of the database management system is often compared and analyzed with the preset operation abnormality coefficient reference threshold. The comparison and analysis results are as follows: If the operation abnormality coefficient is greater than or equal to the preset operation abnormality coefficient reference threshold, it is judged that the operation process of the database management system is in an abnormal operation state, and a high-risk abnormal operation signal is generated; If the operation abnormality coefficient is less than a preset operation abnormality coefficient reference threshold, it is judged that the operation process of the database management system is in a stable operation state, and a low-risk abnormal operation signal is generated.

3. A method for constructing a membrane material database according to claim 2, characterized in that: When a high-risk abnormal operation signal is generated within a fixed detection window during operation of the database management system, the operation abnormality coefficient generated within the fixed detection window during operation of the database management system is continuously obtained to establish an analysis set, and the operation abnormality coefficient in the analysis set is compared and analyzed with the first gradient reference threshold, the second gradient reference threshold, and the operation abnormality coefficient reference, wherein the second gradient reference threshold is greater than the first gradient reference threshold, and the first gradient reference threshold is greater than the operation abnormality coefficient reference threshold, and the second gradient reference threshold, the first gradient reference threshold, and the abnormality coefficient reference threshold are compared and analyzed with the operation abnormality coefficient, and the operation abnormality coefficient that is less than the first gradient reference threshold and greater than or equal to the operation abnormality coefficient reference threshold is calibrated as Za, the operation abnormality coefficient that is less than the second gradient reference threshold and greater than or equal to the first gradient reference threshold is calibrated as Zb, and the operation abnormality coefficient that is greater than or equal to the second gradient reference threshold is calibrated as Zc; Comprehensively analyze Za, Zb, and Zc to generate the abnormal operation coefficient Abnormal OP , based on the formula: Abnormal OP =A1·∑(Za)+A2·∑(Zb)+A3·∑(Zc), where A1, A2, A3 are preset proportional coefficients of Za, Zb, Zc respectively, and A1, A2, A3 are all greater than 0.

4. A method for constructing a membrane material database according to claim 3, characterized in that: The abnormal operation coefficient generated by further analyzing the operation process of the database management system is compared and analyzed with the preset first abnormal operation coefficient reference threshold and the second abnormal operation coefficient reference threshold to determine the risk level of abnormal operation. The comparison and analysis results are as follows: If the abnormal operation coefficient is less than the first abnormal operation coefficient reference threshold, the risk level of the abnormal operation is classified as a low-risk abnormal operation state; If the abnormal operation coefficient is greater than or equal to the first abnormal operation coefficient reference threshold and less than the second abnormal operation coefficient reference threshold, the risk level of the abnormal operation is classified as a medium-risk abnormal operation state; If the abnormal operation coefficient is greater than or equal to the second abnormal operation coefficient reference threshold, the risk level of the abnormal operation is classified as a high-risk abnormal operation state.

5. A method for constructing a membrane material database according to claim 1, characterized in that: The logic for generating the connection pool saturation index by performing an exception analysis on the connection pool saturation information is as follows: In the detection window, collect real-time parameters related to the use of connection pool resources, including the number of active connections in the connection pool C active (t), the maximum number of available connections in the connection pool C max , the number of requests waiting to obtain a connection R wait (t) and the time T that the connection is maintained hol (t); Based on the collected parameters, the connection pool utilization and connection pool waiting rate at each time point are defined. Taking into account the influence of all factors, an exponential decay model is used to give weights to different time points. The calculation expression is as follows: In the formula, U(t) represents the connection pool utilization, W(t) represents the connection pool waiting rate, e represents the natural base, and α represents the exponential decay factor. In the calculation expression of the connection pool utilization U(t), α is used to control the connection holding time T hol The impact of (t) on the connection pool utilization U(t). In the calculation expression of the connection pool waiting rate W(t), α is used to control the holding time T hold (t) Impact on connection pool waiting rate; The dynamic weight factor γ(t) is introduced to capture the dynamic fluctuations of connection pool resources in different time periods and to adjust the impact of different time periods. The calculation expression is as follows: The severity of connection pool saturation is quantified by the abnormality, which combines the connection pool utilization and the connection pool waiting rate. The calculation expression is as follows: E(t) = γ(t)·(U(t)+W(t)) μ , where E(t) represents the abnormality, and μ represents the adjustment coefficient, which is used to balance the influence of utilization rate and waiting rate; Aggregate the abnormality values ​​at all time points and calculate the overall connection pool saturation index. The calculation expression is as follows: In the formula, I pool Indicates the connection pool saturation index, n is the total number of time points in the detection window.

6. A method for constructing a membrane material database according to claim 1, characterized in that: The logic of performing exception analysis on deadlock recovery information to generate a deadlock recovery index is as follows: In the detection window, we first collect multiple dimensions of information related to deadlock during the operation of the database management system, including the deadlock occurrence frequency D freq , the total number of transaction rollbacks caused by deadlock T rollback , the total waiting time of all transactions during deadlock recovery T Waittime , the number of transactions delayed or failed due to deadlock T impact And the deadlock recovery delay R latency ; Deadlock recovery efficiency depends not only on the deadlock frequency freq , and the number of transactions delayed or failed due to deadlocks T impact , the total waiting time of all transactions during deadlock recovery T Waittime and lock recovery delay R latency Therefore, a deadlock impact factor is constructed and the calculation expression is as follows: Where, L impact Represents the deadlock impact factor; To further quantify the difficulty and complexity of deadlock recovery, a deadlock complexity factor is defined. Combining the number of transactions involved in the deadlock and the number of resource types involved, the calculation expression is as follows: L complexity =ln(1+N txn ·N res ), N txn Indicates the number of transactions competing for resources in a deadlock event, representing the complexity of the deadlock, N res Indicates the number of resource types involved in the deadlock; Finally, the deadlock impact factor L impact and the deadlock complexity factor L complexity Combined, the deadlock recovery index is generated, and the calculation expression is as follows: Where DR represents the deadlock recovery index.

7. A system for constructing a membrane material database, used to implement the method for constructing a membrane material database according to any one of claims 1 to 6, characterized in that: It includes data collection and preprocessing module, anomaly detection and identification module, machine learning prediction and status assessment module, and anomaly classification and response strategy module; The data collection and preprocessing module collects multi-dimensional parameter information generated by the system in real time during the operation of the database management system, cleans and preprocesses the collected raw data, and standardizes different types of data to ensure data consistency and validity; The anomaly detection and identification module performs preliminary anomaly detection on the preprocessed operating parameters, identifies abnormal points in the operating state, and determines whether the operating parameters deviate from the normal operating range; The machine learning prediction and status assessment module inputs the parameter information after abnormal analysis into the pre-learned machine learning model and determines the current operating status of the system based on the current parameter change trend; The abnormal classification and response strategy module further comprehensively analyzes the database management system with abnormal operation status, further divides the abnormal operation status of the data management library, and takes different response measures for different abnormal operation status based on the division results; The multi-dimensional parameter information generated during the operation of the database management system includes connection pool saturation information and deadlock recovery information. The connection pool saturation information reflects the resource usage of the connection pool in the database management system. The deadlock recovery information refers to how the database system detects, handles and recovers deadlocks when concurrent transactions are running. After obtaining the connection pool saturation information and deadlock recovery information generated during the operation of the database management system, the connection pool saturation information and the deadlock recovery information are analyzed and processed abnormally, and a connection pool saturation index and a deadlock recovery index are generated respectively. The operation process of the database management system is intelligently judged through the connection pool saturation index and the deadlock recovery index; After obtaining the connection pool saturation index and deadlock recovery index generated during the operation of the database management system, the connection pool saturation index and the deadlock recovery index are input into a pre-learned machine learning model, and an operation anomaly coefficient is generated by the machine learning model. The operation process of the database management system is judged by the operation anomaly coefficient.

Images