WO2021073343A1 - 通信系统的故障根因分析方法、装置、系统和计算机存储介质 - Google Patents

通信系统的故障根因分析方法、装置、系统和计算机存储介质 Download PDF

Info

Publication number
WO2021073343A1
WO2021073343A1 PCT/CN2020/115602 CN2020115602W WO2021073343A1 WO 2021073343 A1 WO2021073343 A1 WO 2021073343A1 CN 2020115602 W CN2020115602 W CN 2020115602W WO 2021073343 A1 WO2021073343 A1 WO 2021073343A1
Authority
WO
WIPO (PCT)
Prior art keywords
variable
communication system
real
observed
variables
Prior art date
Application number
PCT/CN2020/115602
Other languages
English (en)
French (fr)
Inventor
韩静
张百胜
陈力
刘建伟
董辛酉
杨帆
曹亮
郁枫
李维杨
王德政
Original Assignee
中兴通讯股份有限公司
清华大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中兴通讯股份有限公司, 清华大学 filed Critical 中兴通讯股份有限公司
Priority to EP20877967.8A priority Critical patent/EP4033700A4/en
Publication of WO2021073343A1 publication Critical patent/WO2021073343A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/06—Management of faults, events, alarms or notifications
    • H04L41/0631—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/06—Management of faults, events, alarms or notifications
    • H04L41/0631—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
    • H04L41/064—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis involving time analysis
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
    • H04L41/5003—Managing SLA; Interaction between SLA and QoS
    • H04L41/5009—Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/04—Processing captured monitoring data, e.g. for logfile generation
    • H04L43/045—Processing captured monitoring data, e.g. for logfile generation for graphical visualisation of monitoring data
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
    • H04L43/091—Measuring contribution of individual network components to actual service level

Definitions

  • the present disclosure relates to the field of communications, and in particular to a method for analyzing the root cause of a failure of a communication system, a device for analyzing the root cause of a failure of a communication system, a system for performing the method of analyzing the root of a failure, and a computer-readable storage medium.
  • KPI Key Performance Indicator
  • the purpose of the present disclosure is to provide a root cause analysis method for a communication system, a root cause analysis device for a communication system, a system for executing the root cause analysis method, and a computer-readable storage medium.
  • the root cause analysis method can be used to efficiently find the cause of the performance degradation of the communication system.
  • a method for analyzing the root cause of a failure of a communication system includes: acquiring the real-time value of the target KPI variable of the communication system and the real-time value of a plurality of observed variables of the communication system; The real-time values of the multiple observation variables are input into the prediction model to obtain the predicted value of the target KPI variable, wherein the prediction model is obtained by training the historical data of the communication system by a predetermined algorithm; The predicted value of the target KPI variable is compared with the real-time value of the target KPI variable; when the performance of the communication system deteriorates, the contribution of each observed variable to the phenomenon of the deterioration of the system performance is calculated, wherein, The performance degradation of the communication system satisfies at least the following predetermined conditions: the predicted value of the target KPI variable differs from the real-time value of the target KPI variable by more than a predetermined value; The candidate root dependent variable for the deterioration of the performance of the communication system, where
  • a system for analyzing the root cause of a failure of a communication system including: an offline device, the offline device stores historical data obtained by training the communication system using a predetermined algorithm A prediction model for predicting KPI variables; an online device used to: monitor the KPI variables of the communication system in real time to obtain real-time KPI variables; compare the predicted KPI variables output by the prediction model with the real-time KPI variables Contrast; when the performance of the communication system deteriorates, calculate the contribution of each observed variable to the phenomenon of the deterioration of the real-time KPI variable, wherein the performance deterioration of the communication system at least satisfies the following predetermined conditions: The difference between the predicted KPI variable and the real-time KPI variable exceeds a predetermined value; the observed variable whose contribution degree is ranked in the top K according to the magnitude is used as the candidate root dependent variable of the real-time KPI variable deterioration, where K is a positive integer.
  • a computer-readable storage medium stores an executable program, and the executable program is used to implement the above-mentioned root cause analysis of the failure when executed by a processor method.
  • a fault root cause analysis device of a communication system which includes a processor and a memory, the memory stores an executable program, and the executable program is used to implement when being executed by the processor.
  • the above-mentioned failure root cause analysis method includes a processor and a memory, the memory stores an executable program, and the executable program is used to implement when being executed by the processor.
  • FIG. 1 is a flowchart of a method for analyzing the root cause of a failure of a communication system provided by the present disclosure
  • FIG. 2 is a schematic diagram of an implementation manner of step S130;
  • Figure 3 is a flowchart of constructing a predictive model
  • FIG. 4 is a block diagram of a system for analyzing the root sound of a failure in a communication system provided by the present disclosure
  • Figure 5 is a comparison diagram of the predicted value and real-time value of the target KPI before and after the communication system is updated
  • Figure 6 is a schematic diagram of real-time monitoring of T 2 statistics
  • Figure 7 is a schematic diagram of the cumulative contribution of each observed variable
  • Figure 8 is the alarm sequence of the contribution degree of each observation variable
  • Fig. 9 is a schematic structural diagram of a fault root cause analysis device of a communication system provided by the present disclosure.
  • a method for analyzing the root cause of a failure in a communication system includes:
  • step S110 obtain the real-time value of the target KPI variable of the communication system and the real-time value of the multiple observed variables of the communication system;
  • step S120 the real-time values of the multiple observed variables are input into a prediction model to obtain the predicted value of the target KPI variable, wherein the prediction model is trained on the historical data of the communication system by a predetermined algorithm Obtained
  • step S130 the obtained predicted value of the target KPI variable is compared with the real-time value of the target KPI variable
  • step S140 when the performance of the communication system deteriorates, the contribution of each observed variable to the phenomenon of the deterioration of the system performance is calculated, wherein the performance deterioration of the communication system at least satisfies the following predetermined conditions: The difference between the predicted value of the target KPI variable and the real-time value of the target KPI variable exceeds a predetermined value;
  • step S150 the observed variable whose contribution degree ranks the top K in terms of magnitude is used as the candidate root dependent variable of the performance degradation of the communication system, where K is a positive integer.
  • the real-time value of the target KPI variable of the communication system and the real-time value of the multiple observed variables of the communication system can be obtained by real-time monitoring of the communication system.
  • a monitoring device can be used to monitor the communication system in real time, and then the real-time value of the target KPI variable and the multiple observations can be obtained through communication between the monitoring device and the device that executes the fault root cause analysis method The real-time value of the variable.
  • the operating state of the communication system is monitored online to obtain the real-time value of the target KPI variable and the real-time value of a plurality of observed variables.
  • the period for collecting the real-time value of the target KPI variable and collecting the real-time value of each observed variable is not particularly limited.
  • the target KPI variable and the observation variable are collected every time T.
  • the time T may be 15 minutes.
  • the historical data of the communication system refers to the data obtained before the communication system is updated, including the historical values of the target KPI variables before the communication system is updated, and the historical values of the observed variables before the communication system is updated. Also, it is necessary to make sure that the communication system is operating normally before the update.
  • the root cause of the fault that makes the communication system performance worse can be quickly determined, without manual troubleshooting, which improves the efficiency of the root cause analysis of the fault.
  • the observed variable with the largest contribution can generally be considered as the root cause of the fault, but the candidate root dependent variables obtained by the method of root cause analysis of the fault provided in the present disclosure are only for analysis by the technicians. Technicians can determine the true root cause variable of the fault based on the contribution of each observed variable.
  • the specific type of the predetermined algorithm is not particularly limited.
  • the predetermined algorithm includes a partial least squares (PLS, Partial least squares regression) algorithm.
  • PLS Partial least squares regression
  • the PLS algorithm is used to communicate The model obtained by training the historical data of the system can predict the target KPI variable more accurately.
  • the method for analyzing the root cause of the fault may further include performing after step S150:
  • step S160 monitor whether the real-time value of each observed variable exceeds a predetermined control limit
  • step S170 when the real-time value of the K candidate root dependent variables exceeds the corresponding control limit, the frequency is greater than the predetermined frequency and/or the time when the real-time value of the K candidate root dependent variables exceeds the corresponding control limit is earlier than other observations When variable, determine K candidate root dependent variables as final root dependent variables.
  • control limit of each observed variable is not specifically limited, and the technician can set the control limit based on experience.
  • the final root dependent variables are also for reference by technicians.
  • the specific value of K is not particularly limited, and the technician can determine the value of K according to his own needs.
  • the fault root cause analysis method further includes:
  • Construct statistics to monitor the operation status of the communication system online wherein the performance deterioration of the communication system also includes the deviation of the statistics of the communication system (denoted as T 2 ) from the statistical threshold of the communication system (denoted by As T 2 limit ).
  • step S130 may include:
  • step S131 a first curve in which the predicted value of the target KPI changes with the observation time is generated, and a second curve in which the real-time value of the target KPI variable changes with the observation time is generated;
  • step S132 the first curve and the second curve are used for comparison to determine the relationship between the predicted value of the target KPI variable and the real-time value of the target KPI variable at each time point.
  • the observed variables can be sorted according to the degree of contribution according to the following method:
  • both the instantaneous contribution degree of the observed variable and the total contribution degree of the observed variable are considered, so that the contribution degree of each observed variable can be sorted objectively and fairly.
  • the failure root cause analysis method provided by the present disclosure, once it is found that the target KPI has deteriorated (that is, the communication system is abnormal), the contribution of the observed variables in the previous period of time to the KPI deterioration is analyzed, and the A part of the variables with a large contribution degree are identified and regarded as the root cause of the failure for the technical personnel to analyze.
  • the prediction model is constructed through the PLS algorithm.
  • the historical data must be preprocessed first, and the observed variables that have less correlation with the target KIP variable are excluded.
  • the steps of constructing the prediction model include:
  • step S210 in the historical data of the communication system, cross-correlation analysis is performed on the historical value of the target KPI variable of the communication system and the historical value of a plurality of observed variables to obtain the correlation with the historical value of the target KPI variable
  • the historical value of the observed variable whose degree exceeds the first predetermined percentage (denoted as E%) is used as the input variable;
  • step S230 the input variable is used as input data, the historical value of the target KPI variable is used as prediction data, and the predetermined algorithm is used for model training to obtain the prediction model.
  • step S210 among a large number of observed variables, the most relevant measured variable to the target KPI variable can be automatically selected as the input data of the PLS algorithm, thereby reducing the amount of calculation when constructing a predictive model.
  • the inventor of the present application found that most of the effective observation variables of the communication system (that is, the observation variables most related to the target KPI variable) have obvious linear correlation characteristics, that is, the communication system
  • the system is a linear system in which the effective variables exceeding the second predetermined percentage (denoted as D%) have linear correlation characteristics.
  • the first predetermined percentage does not exceed the second predetermined percentage (ie, E ⁇ D). In some embodiments, the first predetermined percentage E% is selected from 50% to 70%, and the second predetermined percentage D% is selected from 70% to 90%.
  • step S230 is executed. Specifically, the step of constructing a prediction model further includes the following steps performed before step S230:
  • step S221 the correlation coefficient between each observed variable and the target KPI variable is determined according to the cross-correlation analysis between the historical value of the target KPI variable and the historical value of the multiple observed variables;
  • step S222 the correlation coefficient threshold is calculated through the significance test
  • step S223 it is determined whether the proportion of the observed variables whose correlation coefficient with the target KPI variable exceeds the correlation coefficient threshold value among all the observed variables exceeds the second predetermined percentage;
  • step S230 is executed.
  • step S223 When the judgment result in step S223 is no, it indicates that the communication system is a non-linear system. At this time, other algorithms may be considered to construct a prediction model.
  • the step of calculating the correlation coefficient between each observed variable and the target KPI variable includes:
  • k is the length of two statistical sequences
  • S xi is the standard deviation of x i
  • S xj is the standard deviation of x j
  • n is the sample number
  • the method of test is to construct a T statistic to obtain the threshold value ⁇ 0 of the correlation coefficient;
  • D can be 70%-90%.
  • the variable with E% before the correlation coefficient is selected as the input of the PLS model, where E ⁇ D.
  • M the number of observed variables in normal operation
  • N the number of samples of observed variables
  • Y the number of samples of observed variables
  • the PLS algorithm projects the data matrix [X,Y] into the low-dimensional space created by fewer latent variables [t 1 ,t 2 ,..., t A ]:
  • T [t 1 ,t 2 ,...,x A ] ⁇ R N ⁇ A is the score matrix
  • E is the residual matrix of X
  • F is the residual matrix of Y.
  • T2 statistic is calculated, using Equation (3) T 2 statistic calculated threshold.
  • T 2 x T R ⁇ -1 R T x (2)
  • T is the sample covariance of the modeling score
  • ⁇ is the confidence level
  • N is the number of samples used when using the PLS algorithm to train the prediction model
  • n is the number of samples used when using the PLS algorithm to train the prediction model
  • A is the number of components selected by the PLS algorithm
  • R is the weight matrix
  • the contribution of each observed variable to the target KPI variable variation within a certain period of time is calculated.
  • the degree of contribution of the observed variable x i to the T 2 monitoring index is determined by formula (4):
  • the system includes an offline device 110 and an online device 120.
  • the offline device 110 stores a prediction model for predicting KPI variables obtained by training historical data of the communication system using a predetermined algorithm.
  • the online device 120 is used to perform the following steps:
  • the observed variables ranked in the top K in terms of contribution degree are used as candidate root dependent variables of the real-time KPI variable deterioration, where K is a positive integer.
  • the offline device 110 can construct the prediction model according to the following steps:
  • the historical value of the target KPI variable of the communication system and the historical value of a plurality of observed variables are cross-correlation analysis to obtain that the correlation degree with the historical value of the target KPI variable exceeds the first predetermined
  • the historical value of the percentage of the observed variable is used as the input variable
  • the input variable is used as input data
  • the historical value of the target KPI variable is used as prediction data
  • the partial least squares algorithm is used for model training to obtain the prediction model.
  • the predetermined algorithm is a PLS algorithm.
  • a computer-readable storage medium stores an executable program, and the executable program is used to implement the functions provided by the present disclosure when executed by a processor.
  • the above-mentioned failure root cause analysis method is provided, the computer-readable storage medium stores an executable program, and the executable program is used to implement the functions provided by the present disclosure when executed by a processor.
  • a device for analyzing the root cause of a failure of a communication system includes a processor 901 and a memory 902.
  • the memory 902 stores an executable program. It is used to implement the above-mentioned fault root cause analysis method provided by the present disclosure when executed by the processor 901.
  • the computer-readable storage medium includes volatile and non-volatile, removable, and removable and non-volatile memory devices implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data).
  • the media cannot be removed.
  • Computer-readable storage media include but are not limited to RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage media, Or any other medium that can be used to store desired information and that can be accessed by a computer.
  • the following uses base station data of a certain cell as an analysis case to describe in detail the working principle of the fault root cause analysis method provided in the present disclosure.
  • the following description is only an exemplary explanation of how the failure root cause analysis method provided by the present disclosure works, and is not a real case.
  • the observation data sampling interval is 15 minutes, and a total of 10 days of data are sampled under normal operating conditions. Remove blank data series and constant value data series to obtain a total of 92 observed variable series and 1 KPI variable series;
  • the correlation threshold obtained by the T statistic is 0.24, and the 89/92 variable exceeds the correlation threshold, and the system can be considered to meet the linear characteristic requirements;
  • the real-time monitoring results of normal data 10 days before the upgrade and the data to be analyzed 7 days after the upgrade are shown in Figure 5.
  • the abscissa in Figure 5 represents time information, where 500 represents the first collected data in chronological order. 500 data points, 1000 represents the 1000th data point collected in chronological order, 1500 represents the 1500th data point collected in chronological order, the ordinate is the value of the target KPI, and the indicator line 1 is the prediction line , The indicator line 2 is the actual line: notice that the system performance has declined after the upgrade (960 sampling points), which is reflected in the predicted performance decline.
  • the T 2 monitoring index ie, the T 2 statistic
  • the control Limit in Figure 6, the abscissa represents the sample sequence, 500 represents the 500th data point collected in chronological order, 1000 represents the 1000th data point collected in chronological order, and 1500 represents the sample sequence collected in chronological order
  • the ordinate represents the T 2 statistic value
  • the horizontal line in the chart area is the control limit of the T 2 statistic
  • the curve is the value of the T 2 monitoring index
  • variable 18 has the highest cumulative contribution, so variable 18 is considered to be the root cause of the fault.
  • the contribution of each variable is analyzed separately, and the alarm is issued when the control limit is exceeded.
  • the number of alarms is the most for the 18th variable, and the alarm is issued first, so it can be considered that the 18th The variable caused the propagation of the fault.
  • the predicted value of the target KPI can be obtained by constructing a prediction model.
  • the predicted value of the target KPI variable By comparing the predicted value of the target KPI variable with the real-time value of the target KPI variable, it is possible to quickly determine whether a communication system fails. Then, by calculating the contribution of each observed variable to the deterioration of the target KPI variable, the root cause of the fault that makes the communication system performance worse can be quickly determined, without manual troubleshooting, which improves the efficiency of the root cause analysis of the fault.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)
  • Debugging And Monitoring (AREA)
  • Telephonic Communication Services (AREA)

Abstract

一种通信系统的故障根因分析方法、装置、系统和计算机可读存储介质，方法包括：获取所述通信系统的目标KPI变量的实时值以及所述通信系统的多个观测变量的实时值；将所述多个观测变量的实时值输入预测模型中，以获得目标KPI变量的预测值，其中，所述预测模型由预定算法对所述通信系统的历史数据进行训练所获得；将获得的所述目标KPI变量的预测值与所述目标KPI变量的实时值进行对比；当所述通信系统的性能变差时，计算各个观测变量对所述系统性能变差这一现象的贡献度；将贡献度按照大小排在前K位的观测变量作为所述通信系统的性能变差的备选根因变量，其中，K为正整数。

Description

通信系统的故障根因分析方法、装置、系统和计算机存储介质
相关申请的交叉引用
本申请基于申请号为201910975121.0、申请日为2019年10月14日的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本申请作为参考。
技术领域
本公开涉及通信领域，具体地，涉及一种通信系统的故障根因分析方法、一种通信系统的故障根因分析装置、一种执行该故障根因分析方法的系统、一种计算机可读存储介质。
背景技术
在通信网络的运行维护过程中，不可避免地会出现通信系统的变更。通信系统变更后可能会出现关键性能指标(KPI，Key Performance Indicator)恶化的问题，技术人员需要在大量观测变量中寻找导致通信系统性能下降的原因，效率较低。
因此，如何高效率地寻找通信系统性能下降的原因成为本领域亟待解决的技术问题。
发明内容
本公开的目的在于提供一种通信系统的故障根因分析方法、一种通信系统的故障根因分析装置、一种执行该故障根因分析方法的系统、一种计算机可读存储介质。利用所述故障根因分析方法可高效率地找到通信系统性能下降的原因。
作为本公开的第一个方面，提供一种通信系统的故障根因分析方法，包括：获取所述通信系统的目标KPI变量的实时值以及所述通信系统的多个观测变量的实时值；将所述多个观测变量的实时值输入预测模型中，以获得目标KPI变量的预测值，其中，所述预测模型由预定算法对所述通信系统的历史数据进行训练所获得；将获得的所述目标KPI变量的预测值与所述目标KPI变量的实时值进行对比；当所述通信系统的性能变差时，计算各个观测变量对所述系统性能变差这一现象的贡献度，其中，所述通信系统的性能变差至少满足以下预定条件：所述目标KPI变量的预测值与所述目标KPI变量的实时值相差超过预定值；将贡献度按照大小排在前K位的观测变量作为所述通信系统的性能变差的备选根因变量，其中，K为正整数。
作为本公开的第二个方面，提供一种用于对通信系统进行故障根因分析的系统，包括：离线装置，所述离线装置中存储有利用预定算法对通信系统的历史数据进行训练获得的用于预测KPI变量的预测模型；在线装置，所述在线装置用于：实时监控通信系统的KPI变量，以获得实时KPI变量；将所述预测模型输出的预测KPI变量与所述实时KPI变量进行对比；当所述通信系统的性能变差时，计算各个观 测变量对所述实时KPI变量变差这一现象的贡献度，其中，所述通信系统的性能变差至少满足以下预定条件：所述预测KPI变量与所述实时KPI变量相差超过预定值；将贡献度按照大小排在前K位的观测变量作为所述实时KPI变量变差的备选根因变量，其中，K为正整数。
作为本公开的第三个方面，提供一种计算机可读存储介质，所述计算机可读存储介质存储有可执行程序，所述可执行程序用于在被处理器执行时实现上述故障根因分析方法。
作为本公开的第四个方面，提供一种通信系统的故障根因分析装置，包括处理器和存储器，所述存储器存储有可执行程序，所述可执行程序用于在被处理器执行时实现上述故障根因分析方法。
附图说明
附图是用来提供对本发明的进一步理解，并且构成说明书的一部分，与下面的具体实施方式一起用于解释本发明，但并不构成对本发明的限制。在附图中：
图1是本公开所提供的通信系统的故障根因分析方法的流程图；
图2是步骤S130的一种实施方式的示意图；
图3是构建预测模型的流程图；
图4是本公开所提供的对通信系统进行故障根音分析的系统的模块图；
图5是通信系统更新前和通信系统更新后，目标KPI的预测值以及实时值的对比图；
图6是T 2统计量实时监控示意图；
图7是各观测变量的累积贡献度示意图；
图8是各观测变量贡献度告警序列；
图9是本公开所提供的通信系统的故障根因分析装置的示意性结构图。
附图标记说明
110：离线装置           120：在线装置
901：处理器             902：存储器
具体实施方式
以下结合附图对本发明的具体实施方式进行详细说明。应当理解的是，此处所描述的具体实施方式仅用于说明和解释本发明，并不用于限制本发明。
作为本发明的第一个方面，提供一种通信系统的故障根因分析方法，如图1所示，该故障根因分析方法包括：
在步骤S110中，获取所述通信系统的目标KPI变量的实时值和所述通信系统的多个观测变量的实时值；
在步骤S120中，将所述多个观测变量的实时值输入预测模型中，以获得所述目标KPI变量的预测值，其中，所述预测模型由预定算法对所述通信系统的历史数据进行训练所获得；
在步骤S130中，将获得的所述目标KPI变量的预测值与所述目标KPI变量的实时值进行对比；
在步骤S140中，当所述通信系统的性能变差时，计算各个观测变量对所述系统性能变差这一现象的贡献度，其中，所述通信系统的性能变差至少满足以下预定条件：所述目标KPI变量的预测值与所述目标KPI变量的实时值相差超过预定值；
在步骤S150中，将贡献度按照大小排在前K位的观测变量作为所述通信系统的性能变差的备选根因变量，其中，K为正整数。
在本公开中，对如何获取通信系统的目标KPI变量的实时值和所述通信系统的多个观测变量的实时值不做特殊的限定。例如，可以通过对通信系统进行实时监控的方式获得所述目标KPI变量的实时值和所述多个观测变量的实时值。再例如，可以利用监控装置对所述通信系统进行实时监控，然后通过监控装置与执行所述故障根因分析方法的装置之间进行通信获得所述目标KPI变量的实时值和所述多个观测变量的实时值。在一些实施例中，通过构造霍特林统计量，在线监测所述通信系统运行的状态，以获得所述目标KPI变量的实时值和多个观测变量的实时值。
在本公开中，对采集目标KPI变量的实时值和采集各个观测变量的实时值的周期也不做特殊限定。例如，每隔时间T采集一次目标KPI变量和观测变量。时间间隔T越小、则故障根因分析方法的分析结果越准确。为了在运算量和精确度之间获得平衡，在一些实施例中，所述时间T可以为15min。
此处，所述通信系统的历史数据，是指通信系统更新前获得的数据，包括通信系统更新前的目标KPI变量的历史值、以及通信系统更新前的观测变量的历史值等。并且，需要确定通信系统在更新前是正常运行的。
通过对比目标KPI变量的预测值和目标KPI变量的实时值，可以快速地判断通信系统是否发生故障。然后，通过计算各个观测变量对目标KPI变量变差的贡献度可以快速地确定使得通信系统性能变差的故障根因，无需人工排查，提高了故障根因分析的效率。
需要指出的是，通常可以认为贡献度最大的观测变量为故障根因变量，但是通过本公开所提供的故障根因分析方法所获得的备选根因变量也仅供技术人员进行分析。技术人员可以根据各个观测变量的贡献度，确定真正的故障根因变量。
在本公开中，对所述预定算法的具体类型不做特殊的限定，在一些实施例中，所述预定算法包括偏最小二乘(PLS，Partial least squares regression)算法。对于通信系统而言，相关的观测变量较多，因此通信系统中涉及的数据量也非常庞大，并且，大采样时间的通信系统满足PLS算法的静态特性，在本申请中，使用PLS算法对通信系统的历史数据进行训练获得的模型可以更加精确地预测目标KPI变量。
为了更加精确地确定故障根因，在一些实施例中，所述故障根因分析方法还可包括在步骤S150之后进行的：
在步骤S160中，监控各个观测变量的实时值是否超过预定的控制限；
在步骤S170中，当K个备选根因变量的实时值超出相应的控制限的频率大于预定频率和/或K个备选根因变量的实时值超出相应的控制限的时机早于其他观测变量时，将K个备选根因变量确定为最终根因变量。
在本公开中，对各个观测变量的控制限不做特殊的限定，技术人员可以根据经验来设定所述控制限。当然，所述的最终根因变量，也是供技术人员进行参考的。
在本公开中，对K的具体数值不做特殊的限定，技术人员可以根据自己的需求来确定K的数值。
为了使得预测根因分析结果更加准确，在一些实施例中，所述故障根因分析方法还包括：
构造统计量，在线监测所述通信系统运行的状态，其中，所述通信系统的性能变差还包括所述通信系统的统计量(记作T 2)偏离所述通信系统的统计量阈值(记作T 2 limit)。
为了让维护人员快速、直观地获得目标KPI变量的预测值和目标KPI变量的实时值之间的对比结果，在一些实施例中，如图2所示，步骤S130可以包括：
在步骤S131中，生成所述目标KPI的预测值随观测时间变化的第一曲线，以及生成所述目标KPI变量的实时值随观测时间变化的第二曲线；
在步骤S132中，利用所述第一曲线和所述第二曲线进行对比，以判断所述目标KPI变量的预测值与所述目标KPI变量的实时值在各个时间点的关系。
在本公开中，对如何按照贡献度对各观测变量进行排序不做特殊的规定，可以按照以下方法按照贡献度对各观测变量进行排序：
计算预定时间段内各观测变量在各个观测时刻的瞬时贡献度；
计算各个观测变量在预定时间段内的累积贡献度；
统计各个观测变量在所述预定时间段内的瞬时贡献度超过第一预定贡献度的次数；
确定所述预定时间段内的累积贡献度超过第二预定贡献度的观测变量；
将瞬时贡献度超过第一预定贡献度的次数大于预定次数、且累计贡献度超过所述第二预定贡献度的观测变量进行排序。
在本公开实施例所提供的具体步骤中，既考虑到了观测变量的瞬时贡献度，又考虑到了观测变量的总贡献度，从而可以更加客观公正地对各观测变量的贡献度进行排序。
换言之，在本公开所提供的故障根因分析方法中，一旦发现目标KPI变差(即，通信系统发生异常)，即分析在此之前一段时间内的观测变量对KPI变差的贡献度，筛选出贡献度较大的一部分变量，并将其视为造成故障的根因，以供技术人员分析。
如上文中所述，通过PLS算法构建所述预测模型。为了节约计算量，首先要对历史数据进行预处理，将与所述目标KIP变量相关性较小的观测变量排除。具体地，如图3所示，构建所述预测模型的步骤包 括：
在步骤S210中，对所述通信系统的历史数据中，所述通信系统的目标KPI变量的历史值与多个观测变量的历史值进行互相关分析，以获得与目标KPI变量的历史值的相关度超过第一预定百分比(记作E％)的观测变量的历史值作为输入变量；
在步骤S230中，将所述输入变量作为输入数据，将所述目标KPI变量的历史值作为预测数据，利用所述预定算法进行模型训练，以获得所述预测模型。
通过步骤S210，可以在大量观测变量中，自动选出与目标KPI变量最相关的测变量，作为PLS算法的输入数据，从而降低了构建预测模型时的运算量。
经本申请的发明人研究发现，所述通信系统的有效观测变量(即，与目标KPI变量最相关的观测变量)中，大部分观测变量存在明显的线性相关特性，也就是说，所述通信系统为超过第二预定百分比(记作D％)的有效变量存在线性相关特性的线性系统。
在一些实施例中，所述第一预定百分比不超过所述第二预定百分比(即，E≤D)。在一些实施例中，第一预定百分比E％选取50％至70％，而第二预定百分比D％则选取70％至90％。
PLS算法尤其适用于对线性系统进行预测。因此，作为提升预测结果的准确性，在一些实施例中，在构建预测模型时，可以首先对通信系统是否为线性系统进行分析。只有确认了通信系统为线性系统，方执行步骤S230。具体地，构建预测模型的步骤还包括在步骤S230之前进行的如下步骤：
在步骤S221中，根据所述目标KPI变量的历史值与多个观测变量的历史值之间的互相关分析，确定各个观测变量与所述目标KPI变量之间的相关系数；
在步骤S222中，通过显著性检验计算得到相关系数阈值；
在步骤S223中，判断与所述目标KPI变量的相关系数超过所述相关系数阈值的观测变量在所有观测变量中的比例是否超过所述第二预定百分比；
当判断结果为是时，表明通信系统为线性系统，则执行步骤S230。
当步骤S223中的判断结果为否时，表明通信系统为非线性系统，此时可以考虑利用其它算法构建预测模型。
在本公开中，对如何执行步骤S221没有特殊的限定。
在本发明中，对如何计算各个观测变量与所述目标KPI变量之间的相关系数不做特殊的限定。具体地，计算各个观测变量与所述目标KPI变量之间的相关系数的步骤包括：
通过互相关分析方法计算KPI变量与其他观测变量的相关系数矩阵P和延时矩阵L。即对两个统计序列x i和x j，引入时延l(l∈Z)，且|l|≤l max，利用公式(3)计算一系列相关系数ρ xixj(l)：
Figure PCTCN2020115602-appb-000001
其中，k为两个统计序列的长度；
S xi为x i的标准差；
S xj为x j的标准差；
m为样本序号。
对于公式(1)获得的一系列相关系数ρ xixj(l)取最大值，记做ρ ij，对应的时延为l ij。
通过显著性检验计算相关系数的阈值，检验的方式是通过构造一个T统计量，以得到相关系数的阈值ρ 0；
如果超过D％的有效变量存在明显线性相关特性，则认为所述通信系统的线性特性明显，满足PLS模型的线性特性。一般情况下，D可以取70％-90％。
从大量观测变量在那个选择相关系数前E％的变量作为PLS模型的输入，其中，E≤D。记正常运行的观测变量个数为M，观测变量的样本个数为N，并对数据进行标准化，得到输入矩阵X＝[x 1,x 2,…，x M]∈R N×M，输出向量为Y＝[y 1,y 2,…,y N] T∈R 1×N。一般情况下，E取50％至70％。
使用PLS算法建立输出变量为目标KPI变量、输入变量为所述观测变量的模型，具体地：
PLS算法将数据矩阵[X,Y]投影到较少潜变量[t 1,t 2,…，t A]所张成的低维空间中：
Figure PCTCN2020115602-appb-000002
式中，T＝[t 1,t 2,…，x A]∈R N×A为得分矩阵；
P＝[p 1,p 2,…,p A]∈R M×A为X的载荷矩阵；
Q＝[q 1,q 2,…,q A]∈R M×A为Y的载荷矩阵；
E为X的残差矩阵；
F为Y的残差矩阵。
引入权重矩阵R＝[r 1,r 2,…,r A]∈R M×A，使得T＝XR，进一步可以得到P TR＝R TP，则对Y的预测模型可以写成
Figure PCTCN2020115602-appb-000003
下面介绍如何利用所述预测模型来进行故障根因分析：
如上文中所述，构造T 2统计量在线监测通信系统的运行状态；
将观测变量的实时值输入上述预测模型中，获得目标KPI的预测变量；
如果目标KPI的实时值显著偏离目标KPI的预测变量，且T 2统计量显著偏离T 2 limit阈值，则认为通信系统的性能下降。其中，利用公式(2)计算T2统计量，利用公式(3)计算T 2统计量阈值。
T 2＝x TRΔ -1R Tx        (2)
Figure PCTCN2020115602-appb-000004
其中，
Figure PCTCN2020115602-appb-000005
T为建模得分的样本协方差；
α为置信水平；
N为利用PLS算法训练预测模型时用到的样本个数；
n为利用PLS算法训练预测模型时用到的样本个数样本个数；
A为PLS算法选取的成分个数；
R为权重矩阵。
一旦监测到通信系统的性能下降，则计算一定时间内各个观测变量对目标KPI变量变差的贡献度。采用重构贡献图方法(RBC)，观测变量x i对T 2监测指标的贡献程度。具体地，可以通过公式(4)确定贡献度RBC：
Figure PCTCN2020115602-appb-000006
其中，
Figure PCTCN2020115602-appb-000007
为故障向量。
作为本公开的第二个方面，提供一种用于对通信系统进行故障根因分析的系统，如图4所示，该系统包括离线装置110和在线装置120。
其中，离线装置110中存储有利用预定算法对通信系统的历史数据进行训练获得的用于预测KPI变量的预测模型。
在线装置120用于执行以下步骤：
实时监控通信系统的KPI变量，以获得实时KPI变量；
将所述预测模型输出的预测KPI变量与所述实时KPI变量进行对比；
当所述通信系统的性能变差时，计算各个观测变量对所述实时KPI变量变差这一现象的贡献度，其中，所述通信系统的性能变差至少满足以下预定条件：所述预测KPI变量与所述实时KPI变量相差超过预定值；
将贡献度按照大小排在前K位的观测变量作为所述实时KPI变量变差的备选根因变量，其中，K为 正整数。
在一些实施例中，离线装置110能够按照以下步骤构建所述预测模型：
对所述通信系统的历史数据中，所述通信系统的目标KPI变量的历史值与多个观测变量的历史值进行互相关分析，以获得与目标KPI变量的历史值的相关度超过第一预定百分比的观测变量的历史值作为输入变量；
将所述输入变量作为输入数据，将所述目标KPI变量的历史值作为预测数据，利用偏最小二乘算法进行模型训练，以获得所述预测模型。
在一些实施例中，所述预定算法为PLS算法。
作为本公开的第三个方面，提供一种计算机可读存储介质，所述计算机可读存储介质存储有可执行程序，所述可执行程序用于在被处理器执行时实现本公开所提供的上述故障根因分析方法。
作为本公开的第四个方面，提供一种通信系统的故障根因分析装置，如图9所示，包括处理器901和存储器902，所述存储器902存储有可执行程序，所述可执行程序用于在被处理器901执行时实现本公开所提供的上述故障根因分析方法。
其中，计算机可读存储介质包括在用于存储信息(诸如计算机可读指令、数据结构、程序模块或其他数据)的任何方法或技术中实施的易失性和非易失性、可移除和不可移除介质。计算机可读存储介质包括但不限于RAM、ROM、EEPROM、闪存或其他存储器技术、CD-ROM、数字多功能盘(DVD)或其他光盘存储、磁盒、磁带、磁盘存储或其他磁存储介质、或者可以用于存储期望的信息并且可以被计算机访问的任何其他的介质。
下面以某小区的基站数据作为分析案例，详细描述本公开所提供的故障根因分析方法的工作原理。以下说明仅为示例性解释本公开所提供的故障根因分析方法是如何工作的，并非真实案例。
某小区共201个待观测变量，评价系统性能的有1个目标KPI变量，观测数据采样间隔为15min，正常运行状态下共采样10天数据。去除空白数据序列以及常值数据序列，共得到92个观测变量序列和1个KPI变量序列；
利用互相关分析法进行分析，T统计量得到的相关性阈值为0.24，其中，89/92的变量超过了相关性阈值，可以认为系统满足线性特性要求；
选取相关系数前65％的变量作为PLS算法的输入，得到60个输入变量；
使用PLS算法为目标KPI建立预测模型；
构造T 2监测指标对系统进行实时监控；
升级前10天正常10天正常数据和升级后7天待分析数据的实时监控结果图5所示，其中，图5的横坐标表示的是时间信息，其中，500表示按照时间顺序采集到的第500个数据点，1000表示按照时间顺序采集到的第1000个数据点，1500则表示按照时间顺序采集到的第1500个数据点，纵坐标为目 标KPI的值，其中，指示线1为预测线，指示线2为实际线：注意到升级后(960个采样点)系统性能发生了下降，体现在预测性能下降，如图6所示，T 2监测指标(即，T 2统计量)超过控制限(在图6中，横坐标表示样本序列，500表示按照时间顺序采集到的第500个数据点，1000表示按照时间顺序采集到的第1000个数据点，1500则表示按照时间顺序采集到的第1500个数据点，纵坐标表示的是T 2统计值，图表区域内的水平横线为T 2统计量的控制限，曲线为T 2监测指标的值)，因而需要分析各变量对统计指标的贡献程度。
取分析长度为1天，即96个采样点。如图7所示，所述故障根因分析系统给出的累计贡献率最高的前五个观测变量(即，K为5)分别为：18→27→19→42→79，通过图7可以看出18号变量累计贡献最高，因而认为18号变量是故障根因。另一方面，对每个变量单独进行贡献度分析，超过控制限的发出报警，如图8所示，18号变量同样可以看出其报警次数最多，且最先发出告警，因而可以认为18号变量引起了故障的传播。
在本公开所提供的故障根因分析方法的若干实施例中，通过构建预测模型，可以得到目标KPI的预测值。通过对比目标KPI变量的预测值和目标KPI变量的实时值，可以快速地判断通信系统是否发生故障。然后，通过计算各个观测变量对目标KPI变量变差的贡献度可以快速地确定使得通信系统性能变差的故障根因，无需人工排查，提高了故障根因分析的效率。
可以理解的是，以上实施方式仅仅是为了说明本发明的原理而采用的示例性实施方式，然而本发明并不局限于此。对于本领域内的普通技术人员而言，在不脱离本发明的精神和实质的情况下，可以做出各种变型和改进，这些变型和改进也视为本发明的保护范围。

Claims (14)

  1. 一种通信系统的故障根因分析方法，包括：
    获取所述通信系统的目标KPI变量的实时值以及所述通信系统的多个观测变量的实时值；
    将所述多个观测变量的实时值输入预测模型中，以获得目标KPI变量的预测值，其中，所述预测模型由预定算法对所述通信系统的历史数据进行训练所获得；
    将获得的所述目标KPI变量的预测值与所述目标KPI变量的实时值进行对比；
    当所述通信系统的性能变差时，计算各个观测变量对所述系统性能变差这一现象的贡献度，其中，所述通信系统的性能变差至少满足以下预定条件：所述目标KPI变量的预测值与所述目标KPI变量的实时值相差超过预定值；
    将贡献度按照大小排在前K位的观测变量作为所述通信系统的性能变差的备选根因变量，其中，K为正整数。
  2. 根据权利要求1所述的故障根因分析方法，其中，获取所述通信系统的目标KPI变量的实时值以及所述通信系统的多个观测变量的实时值具体包括：
    构造统计量，在线监测所述通信系统运行的状态，以获得所述通信系统的目标KPI变量的实时值以及所述通信系统的多个观测变量，其中，所述通信系统的性能变差还包括所述通信系统的统计量偏离所述通信系统的统计量阈值。
  3. 根据权利要求1所述的故障根因分析方法，其中，将获得的所述目标KPI变量的预测值与所述目标KPI变量的实时值进行对比，包括：
    生成所述目标KPI变量的预测值随观测时间变化的第一曲线，以及生成所述目标KPI变量的实时值随观测时间变化的第二曲线；
    利用所述第一曲线和所述第二曲线进行对比，以判断所述目标KPI变量的预测值与所述目标KPI变量的实时值在各个时间点的关系。
  4. 根据权利要求1至3中任意一项所述的故障根因分析方法，其中，所述故障根因分析方法还包括将贡献度按照大小排在前K位的观测变量作为所述通信系统的性能变差的备选根因变量的步骤之前进行的：
    计算预定时间段内各观测变量在各个观测时刻的瞬时贡献度；
    计算各个观测变量在预定时间段内的累积贡献度；
    统计各个观测变量在所述预定时间段内的瞬时贡献度超过第一预定贡献度的次数；
    确定所述预定时间段内的累积贡献度超过第二预定贡献度的观测变量；
    将瞬时贡献度超过第一预定贡献度的次数大于预定次数、且累计贡献度超过所述第二预定贡献度的 观测变量进行排序。
  5. 根据权利要求1至4中任意一项所述的故障根因分析方法，其中，所述预定算法为偏最小二乘算法。
  6. 根据权利要求5所述的故障根因分析方法，其中，所述预测模型通过以下步骤构建：
    对所述通信系统的历史数据中，所述通信系统的目标KPI变量的历史值与多个观测变量的历史值进行互相关分析，以获得与目标KPI变量的历史值的相关度超过第一预定百分比的观测变量的历史值作为输入变量；
    将所述输入变量作为输入数据，将所述目标KPI变量的历史值作为预测数据，利用所述预定算法进行模型训练，以获得所述预测模型。
  7. 根据权利要求6所述的故障根因分析方法，其中，所述通信系统为超过第二预定百分比的观测变量存在线性相关特性的线性系统，所述第一预定百分比不超过所述第二预定百分比。
  8. 根据权利要求7所述的故障根因分析方法，其中，构建所述预测模型还包括在将所述输入变量作为输入数据，将所述目标KPI变量作为预测数据，利用预定算法进行模型训练，以获得所述预测模型的步骤之前进行的以下步骤：
    根据所述目标KPI变量的历史值与多个观测变量的历史值之间的互相关分析，确定各个观测变量与所述目标KPI变量之间的相关系数；
    通过显著性检验计算得到相关系数阈值；
    判断与所述目标KPI变量的相关系数超过所述相关系数阈值的观测变量在所有观测变量中的比例是否超过所述第二预定百分比；
    当判断结果为是时，则执行所述将所述输入变量作为输入数据，将所述目标KPI变量作为预测数据，利用偏最小二乘算法进行模型训练，以获得预测模型的步骤。
  9. 根据权利要求1至3中任意一项所述的故障根因分析方法，其中，所述故障根因分析方法还包括在将贡献度按照大小排在前K位的观测变量作为所述通信系统的性能变差的备选根因变量的步骤之后进行的：
    监控各个观测变量的实时值是否超过预定的控制限；
    当K个备选根因变量的实时值超出相应的控制限的频率大于预定频率，以及K个备选根因变量的实时值超出相应的控制限的时机早于其他观测变量时，将K个备选根因变量确定为最终根因变量。
  10. 根据权利要求1至3中任意一项所述的故障根因分析方法，其中，所述故障根因分析方法还包括在将贡献度按照大小排在前K位的观测变量作为所述通信系统的性能变差的备选根因变量的步骤之后进行的：
    监控各个观测变量的实时值是否超过预定的控制限；
    当K个备选根因变量的实时值超出相应的控制限的频率大于预定频率，或者K个备选根因变量的实时值超出相应的控制限的时机早于其他观测变量时，将K个备选根因变量确定为最终根因变量。
  11. 一种用于对通信系统进行故障根因分析的系统，包括：
    离线装置，所述离线装置中存储有利用预定算法对通信系统的历史数据进行训练获得的用于预测KPI变量的预测模型；
    在线装置，所述在线装置用于：
    实时监控通信系统的KPI变量，以获得实时KPI变量；
    将所述预测模型输出的预测KPI变量与所述实时KPI变量进行对比；
    当所述通信系统的性能变差时，计算各个观测变量对所述实时KPI变量变差这一现象的贡献度，其中，所述通信系统的性能变差至少满足以下预定条件：所述预测KPI变量与所述实时KPI变量相差超过预定值；
    将贡献度按照大小排在前K位的观测变量作为所述实时KPI变量变差的备选根因变量，其中，K为正整数。
  12. 根据权利要求11所述的系统，其中，所述预定算法为偏最小二乘算法。
  13. 一种计算机可读存储介质，所述计算机可读存储介质存储有可执行程序，所述可执行程序用于在被处理器执行时实现权利要求1至10中任意一项所述的故障根因分析方法。
  14. 一种通信系统的故障根因分析装置，包括处理器和存储器，所述存储器存储有可执行程序，所述可执行程序用于在被所述处理器执行时实现权利要求1至10中任意一项所述的故障根因分析方法。
PCT/CN2020/115602 2019-10-14 2020-09-16 通信系统的故障根因分析方法、装置、系统和计算机存储介质 WO2021073343A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP20877967.8A EP4033700A4 (en) 2019-10-14 2020-09-16 METHOD AND DEVICE FOR ANALYZING THE TRUE CAUSE OF FAILURE OF A COMMUNICATION SYSTEM, SYSTEM AND COMPUTER STORAGE MEDIUM

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910975121.0A CN110635952B (zh) 2019-10-14 2019-10-14 通信系统的故障根因分析方法、系统和计算机存储介质
CN201910975121.0 2019-10-14

Publications (1)

Publication Number Publication Date
WO2021073343A1 true WO2021073343A1 (zh) 2021-04-22

Family

ID=68974878

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/115602 WO2021073343A1 (zh) 2019-10-14 2020-09-16 通信系统的故障根因分析方法、装置、系统和计算机存储介质

Country Status (3)

Country Link
EP (1) EP4033700A4 (zh)
CN (1) CN110635952B (zh)
WO (1) WO2021073343A1 (zh)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110635952B (zh) * 2019-10-14 2021-03-16 中兴通讯股份有限公司 通信系统的故障根因分析方法、系统和计算机存储介质
CN110888755B (zh) * 2019-11-15 2023-04-11 亚信科技（中国）有限公司 一种微服务系统异常根因节点的查找方法及装置
CN111338921A (zh) * 2020-02-21 2020-06-26 深圳壹账通智能科技有限公司 一种系统性能预测方法、装置、计算机设备和存储介质
CN113422690A (zh) * 2020-03-02 2021-09-21 烽火通信科技股份有限公司 一种业务质量劣化预测方法及系统
CN113518373A (zh) * 2020-04-10 2021-10-19 中兴通讯股份有限公司 识别问题小区的方法、电子设备、计算机可读介质
CN113556258B (zh) * 2020-04-24 2022-12-27 西安华为技术有限公司 一种异常检测方法及装置
CN111865407B (zh) * 2020-06-11 2021-11-30 烽火通信科技股份有限公司 光通道性能劣化智能预警方法、装置、设备及存储介质
CN112087334B (zh) * 2020-09-09 2022-10-18 中移（杭州）信息技术有限公司 告警根因分析方法、电子设备和存储介质
CN112929363B (zh) * 2021-02-04 2022-05-17 北京字跳网络技术有限公司 视频领域性能参数异常的根因分析方法及设备
CN113282884B (zh) * 2021-04-28 2023-09-26 沈阳航空航天大学 通用根因分析方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130318011A1 (en) * 2012-05-24 2013-11-28 Michael Jeffrey Jones Method for Detecting Anomalies in Multivariate Time Series Data
CN104537260A (zh) * 2015-01-14 2015-04-22 清华大学 基于缓慢特征回归的动态软测量方法和系统
CN109063886A (zh) * 2018-06-12 2018-12-21 阿里巴巴集团控股有限公司 一种异常检测方法、装置以及设备
CN109710983A (zh) * 2018-12-04 2019-05-03 北京大学 一种基于关键性能指标的柴油机气缸分层故障诊断方法
CN110635952A (zh) * 2019-10-14 2019-12-31 中兴通讯股份有限公司 通信系统的故障根因分析方法、系统和计算机存储介质

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9401851B2 (en) * 2014-03-28 2016-07-26 Verizon Patent And Licensing Inc. Network management system
US10031510B2 (en) * 2015-05-01 2018-07-24 Aspen Technology, Inc. Computer system and method for causality analysis using hybrid first-principles and inferential model
US20170364819A1 (en) * 2016-06-17 2017-12-21 Futurewei Technologies, Inc. Root cause analysis in a communication network via probabilistic network structure
CN110326257B (zh) * 2017-03-01 2022-11-29 瑞典爱立信有限公司 使用人工生命预报关键性能指示符的方法和设备
CN109884892B (zh) * 2019-02-28 2022-08-02 杭州电子科技大学 基于交叉相关时滞灰色关联分析的流程工业系统预测模型

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130318011A1 (en) * 2012-05-24 2013-11-28 Michael Jeffrey Jones Method for Detecting Anomalies in Multivariate Time Series Data
CN104537260A (zh) * 2015-01-14 2015-04-22 清华大学 基于缓慢特征回归的动态软测量方法和系统
CN109063886A (zh) * 2018-06-12 2018-12-21 阿里巴巴集团控股有限公司 一种异常检测方法、装置以及设备
CN109710983A (zh) * 2018-12-04 2019-05-03 北京大学 一种基于关键性能指标的柴油机气缸分层故障诊断方法
CN110635952A (zh) * 2019-10-14 2019-12-31 中兴通讯股份有限公司 通信系统的故障根因分析方法、系统和计算机存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
GE ZHIQIANG: "Distributed predictive modeling framework for prediction and diagnosis of key performance index in plant-wide processes", JOURNAL OF PROCESS CONTROL, vol. 65, 19 October 2017 (2017-10-19), pages 107 - 117, XP055802210, DOI: 10.1016/j.jprocont.2017.08.010 *

Also Published As

Publication number Publication date
CN110635952A (zh) 2019-12-31
EP4033700A4 (en) 2022-11-30
EP4033700A1 (en) 2022-07-27
CN110635952B (zh) 2021-03-16

Similar Documents

Publication Publication Date Title
WO2021073343A1 (zh) 通信系统的故障根因分析方法、装置、系统和计算机存储介质
CN106951984B (zh) 一种系统健康度动态分析预测方法及装置
CN106886481B (zh) 一种系统健康度静态分析预测方法及装置
US7243049B1 (en) Method for modeling system performance
US8676727B2 (en) Abnormality diagnosis filter generator
US20220341996A1 (en) Method for predicting faults in power pack of complex equipment based on a hybrid prediction model
Ottaviani et al. Multiple linear regression model for improved project cost forecasting
KR101776350B1 (ko) 압축기를 진단하는 방법 및 시스템
KR100960939B1 (ko) 최소 절단 집합과 퍼지 전문가 시스템을 이용한 시스템의고장 모드 및 임계 분석 장치 및 그 방법
US7617313B1 (en) Metric transport and database load
CN114997607A (zh) 一种基于工程检测数据的异常评估预警方法及系统
JP2009086896A (ja) コンピュータの障害予測システムおよび障害予測方法
CN114037140A (zh) 预测模型训练、数据预测方法、装置、设备及存储介质
KR101851367B1 (ko) 신용도를 평가하는 방법, 장치 및 컴퓨터 판독 가능한 기록 매체
CN114155048A (zh) 关联业务预测的方法、装置、电子设备及存储介质
Siliverstovs et al. Forecasting industrial production with linear, nonlinear, and structural change models
KR101884908B1 (ko) 빅데이터 분석 기반 신뢰도 예측 장치
Moat et al. Survival analysis and predictive maintenance models for non-sensored assets in facilities management
CN116957534A (zh) 一种预测智能电表更换数量的方法
KR101884907B1 (ko) 빅데이터 분석 기반 신뢰도 예측 방법
CN115688016A (zh) 一种用于大型数据库智能运维的一体化方法
CN115392056A (zh) 一种高压架空输电线路运行状态监测及预警方法及装置
Parvez et al. An association rule mining approach to predict alarm events in industrial alarm floods
KR101469081B1 (ko) 실시간 성적 분포 및 등급컷 추정 방법
Wang Modelling condition monitoring inspection using the delay-time concept

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20877967

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2020877967

Country of ref document: EP

Effective date: 20220421