A remote online monitoring method and system for communication equipment failure
By performing dimensionality reduction and difference analysis on communication device data, and adjusting cluster center splitting in the ISODATA algorithm, the problem of inaccurate clustering results in the existing technology is solved, and more accurate communication device failure detection is achieved.
Patent Information
- Application Number
- CN202410212975.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-27
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-02-27
AI Technical Summary
In the fault detection of communication equipment, the existing ISODATA algorithms have great differences and different importance when the equipment is running normally, resulting in unreasonable splitting of the cluster center and inaccurate clustering results.
By obtaining multi-dimensional data of communication devices, dimensionality reduction is performed to obtain principal component data and linear combination coefficients, calculate the matching principal component and difference, adjust the clustering center split, and use the corrected discrete degree to perform clustering to ensure the clustering results accuracy.
It effectively avoids the influence of redundant information on clustering results, improves the accuracy of clustering centers and the accuracy of clustering results, and enhances the reliability of communication equipment failure detection.
Smart Images

Figure CN117786399B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of communication equipment data measurement, and in particular to a method and system for remote online monitoring of communication equipment faults. Background Art
[0002] Through remote online monitoring of various data of communication equipment, it can help monitor and diagnose communication equipment failures to ensure its normal operation. After obtaining data from sensor devices and communication equipment itself, timely detection and problem diagnosis can be carried out. The remote monitoring system can help avoid or reduce service interruptions, ensure the security and stability of the system, and thus improve user satisfaction.
[0003] In the prior art, the ISODATA algorithm can be used to perform cluster analysis on multiple communication devices, and equipment failures can be detected based on abnormal clusters in the clustering results. In the traditional ISODATA algorithm, each cluster split is determined based on the discrete degree of each dimensional data in the sample set. However, in the actual clustering process, the data of each dimensionality will have obvious differences when the equipment is operating normally, and the importance of each dimensional data in different devices is different, which will cause unreasonable splitting of the cluster center and inaccurate clustering results. Summary of the invention
[0004] In order to solve the technical problem that in the ISODATA algorithm, the data of each dimension will have obvious differences when the equipment is operating normally and the importance of each dimension data in different equipment is different, which will cause unreasonable splitting of cluster centers and thus inaccurate clustering results, the purpose of the present invention is to provide a remote online monitoring method and system for communication equipment faults, and the technical solutions adopted are as follows:
[0005] A method for remote online monitoring of communication equipment failures, the method comprising:
[0006] Obtain communication data of each dimension of each communication device; the dimensions include at least: power load, CPU utilization, memory usage, and traffic data;
[0007] Performing dimensionality reduction on the multi-dimensional communication data of each communication device to obtain the principal component data of each communication device, the linear combination coefficient corresponding to the principal component data, and the variance contribution rate; obtaining the matching principal component between any two communication devices according to the correlation coefficient between the linear combination coefficients of the principal component data in different communication devices;
[0008] When clustering is performed using the ISODATA algorithm, all communication devices are initially clustered using preset initial parameters to obtain initial sample clusters; the difference between the matching principal components of any two communication devices is obtained; the degree of dispersion of the initial sample cluster under any matching principal component is obtained based on the difference between the corresponding variance contribution rate of any matching principal component between any two communication devices in each of the initial sample clusters and the matching principal component; the cluster center of the initial sample cluster is split according to the degree of dispersion to obtain the cluster center point after the split;
[0009] Obtaining clustering results; and performing equipment fault detection according to the clustering results.
[0010] Furthermore, the matching main component acquisition method includes:
[0011] The correlation coefficient between each principal component data of a communication device and the linear combination coefficient of all principal component data of any other communication device is calculated, and the two principal component data with the largest correlation coefficient are used as matching principal components of any two communication devices.
[0012] Furthermore, the correlation coefficient is a Pearson correlation coefficient.
[0013] Furthermore, the method of obtaining the difference includes:
[0014] Calculate the square of the difference between the first-order differences of the corresponding data at each moment under any matching principal component in any two communication devices as the first-order difference similarity, and perform negative correlation mapping and normalization processing on the first-order difference similarity to obtain the first-order difference similarity of the corresponding data at each moment under any matching principal component in any two communication devices; the first-order difference at each moment is the difference between the principal component data at the next moment and the principal component data at the corresponding moment;
[0015] Normalize the sequence number of each moment within the preset time to obtain the weight coefficient;
[0016] The weight coefficient is used to weight the first-order difference similarity of the data corresponding to each moment, so as to obtain the similarity of the data corresponding to each moment in any matching principal component of any two communication devices within a preset time;
[0017] The difference between the data means of any matching principal components in any two communication devices is taken as the degree of data difference;
[0018] The difference between any matching main components in any two communication devices is obtained according to the data difference degree of any matching main components in any two communication devices within a preset time and the similarity degree of corresponding data at each moment.
[0019] Furthermore, the method of obtaining the difference according to the similarity degree and the data difference degree includes:
[0020] The difference is obtained according to the difference calculation formula, and the difference calculation formula is as follows:
[0021] ;
[0022] In the formula, Indicates A communication device and The communication device is The differences between the matching principal components; Indicates The first communication device Matching principal components; Indicates The first communication device Matching principal components; Indicates The first communication device The data means within the matching principal components; Indicates The first communication device The data means within the matching principal components; Indicates The current moment of the matching principal component within the preset time; Indicates The time sequence number of the matching principal component within the preset time; Indicates The first communication device Matching principal components The first-order difference of the data corresponding to each moment; Indicates The first communication device Matching principal components The first-order difference of the data corresponding to each moment; represents the normalization function; Represents an exponential function with a natural constant as its base.
[0023] Furthermore, the discrete degree acquisition method includes:
[0024] The discrete degree is obtained according to the discrete degree calculation formula, and the discrete degree calculation formula is as follows:
[0025] ;
[0026] In the formula, Indicates that each initial sample cluster is The degree of dispersion under the matching principal components; Indicates the number of communication devices included in the initial sample cluster; Indicates the first The first communication device The corresponding variance contribution rate of the matching principal components; Indicates the first The first communication device The corresponding variance contribution rate of the matching principal components; Indicates A communication device and The communication device is The differences between the matching principal components; Indicates A communication device and The communication device is The initial degree of dispersion between the matching principal components.
[0027] Furthermore, the cluster center points after splitting are obtained, including:
[0028] Averaging each corresponding principal component to be analyzed in all communication devices in the initial sample cluster to obtain an initial cluster center;
[0029] The cluster center adjustment vector is obtained by taking the square root of the discrete degree of each corresponding principal component to be analyzed in all communication devices in each of the initial sample clusters; a deviation coefficient is preset, and the product of the deviation coefficient and the cluster center adjustment vector is used as the cluster center adjustment coefficient; and the initial cluster center is split according to the cluster center adjustment coefficient.
[0030] Further, performing equipment fault detection according to the clustering result includes:
[0031] According to the clustering results, the class with the least number of samples among all clusters is selected as the suspected faulty communication equipment.
[0032] Furthermore, the initial parameters include:
[0033] The initial number of cluster centers ; The minimum number of samples in each initial sample cluster ; The degree of dispersion of sample distribution in an initial sample cluster ; The threshold for merging initial sample clusters ; The maximum number of pairs of cluster centers that can be merged in one iteration ; Maximum number of iterations ; The sample is the communication data in each of the communication devices.
[0034] A remote online monitoring system for communication equipment faults, the system comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of a remote online monitoring method for communication equipment faults when executing the computer program.
[0035] The present invention has the following beneficial effects:
[0036] The present invention obtains communication data of each dimension of each communication device; performs dimensionality reduction on the communication data of each dimension, removes redundant information that may exist in the multi-dimensional communication data, and ensures that the information loss is not too much; uses the linear combination coefficients of the principal component data in different communication devices to match, obtains all matching principal components between any two communication devices, and the matching principal components can determine which principal component data of different communication devices have similarities; obtains an initial sample cluster, and analyzes the degree of discreteness between the principal component data corresponding to each communication device in the initial sample cluster; calculates the degree of similarity and the degree of difference of each data between the matching principal components of any two communication devices, and then obtains the difference between the matching principal components of any two communication devices, and the calculation of the degree of similarity takes into account the influence of the change trend of each data between the matching principal components of any two communication devices within a preset time on the difference; uses the obtained difference to obtain the degree of discreteness of the initial sample cluster under any matching principal component, and the degree of discreteness takes into account the change trend of the data, and corrects the degree of discreteness in the initial sample cluster; uses the degree of discreteness to bring into the ISODATA algorithm for clustering, obtains the clustering result, and performs equipment fault detection. The present invention reduces the dimensions of the communication data of each dimension of the communication device, avoiding the influence of redundant information on the clustering results. In the ISODATA algorithm clustering process, the influence of the change trend of the data on the discrete degree of the cluster cluster is taken into account, so that the cluster center point after the split is more accurate and the clustering result is more precise. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0038] Figure 1 A flow chart of a method for remote online monitoring of communication equipment failures provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0039] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a method and system for remote online monitoring of communication equipment faults proposed by the present invention, its specific implementation, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, any two "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0040] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0041] The specific scheme of a remote online monitoring method and system for communication equipment failure provided by the present invention is described in detail below with reference to the accompanying drawings.
[0042] See also Figure 1 , which shows a flow chart of a method for remote online monitoring of communication equipment faults provided by an embodiment of the present invention, the method comprising:
[0043] Step S1: Obtain communication data of each dimension of each communication device; the dimensions include at least: power load, CPU utilization, memory usage, and traffic data.
[0044] The embodiment of the present invention provides a remote online monitoring method for communication equipment faults. In order to detect the faults of communication equipment, it is necessary to first obtain the communication data of each dimension of each communication equipment. The communication data includes the communication data of multiple communication equipment. With the help of sensors and the monitoring system of the communication equipment itself, the communication data of multiple communication equipment can be obtained. In one embodiment of the present invention, the communication data of the dimensions of power load, CPU utilization, memory usage, flow data, etc. are collected respectively, and the communication data of each dimension of each communication equipment collected is preprocessed.
[0045] In one embodiment of the present invention, the communication data of multiple communication devices in various dimensions are acquired, and the data of 1 minute forward from the current time is analyzed, and the sampling interval is 1 second; the total number of communication devices detected is set to , and remove the dimension effect of all sampled data. It should be noted that in other embodiments of the present invention, the data sampling step can be set by specific implementation personnel according to specific implementation scenarios, and is not limited here.
[0046] Step S2: Reduce the dimensionality of the multidimensional communication data of each communication device to obtain the principal component data of each communication device, the linear combination coefficient corresponding to the principal component data, and the variance contribution rate; obtain the matching principal component between any two communication devices based on the correlation coefficient between the linear combination coefficients of the principal component data in different communication devices.
[0047] Considering that there may be a large amount of redundant information in the multi-dimensional communication data obtained in step S1, that is, there is a strong correlation between the communication data of certain dimensions, in order to remove redundant information and ensure that the information loss is not too much, the embodiment of the present invention needs to perform data dimensionality reduction before analyzing and processing the communication data.
[0048] In one embodiment of the present invention, the PCA principal component analysis method is used to reduce the dimension of the communication data of each dimension of each communication device to obtain the principal component data of each communication device required for subsequent processing and the linear combination coefficient and the corresponding variance contribution rate of each principal component data. The specific calculation process is a technical means well known to those skilled in the art and will not be elaborated here.
[0049] It should be noted that in other embodiments of the present invention, the LDA linear discriminant analysis method can also be used to reduce the dimension of the communication data to obtain the principal component data of each communication device and the linear combination coefficient and the corresponding variance contribution rate corresponding to each principal component data.
[0050] Only one embodiment of the present invention is briefly described. In order to reduce the redundancy of data, dimension reduction is performed on each communication device, and the principal component data of the same number before the eigenvalue is selected for subsequent calculation. The method for obtaining the principal component data is as follows:
[0051] In one embodiment of the present invention, the principal component data acquisition method includes:
[0052] All dimensional data of each communication device are reduced in dimension using PCA principal component analysis, and the same preset number of principal components before the eigenvalues are selected in each communication device so that the sum of the corresponding variance contribution rates of the same preset number of principal components in each communication device is greater than the preset first threshold; the same preset number of principal components are used as principal component data. In one embodiment of the present invention, the preset first threshold is set to 85%, that is, the first z principal components whose corresponding variance contribution rates of the principal components in each communication device are greater than 85% are used as principal component data.
[0053] It should be noted that the PCA principal component analysis algorithm is a technical means well known to those skilled in the art, and the preset first threshold and the preset number of principal component data can be set by the implementers according to the specific implementation scenario and are not limited here.
[0054] The principal component data and their corresponding linear combination coefficients in different communication devices are not the same, and each principal component data represents a different proportion of the original communication data of the communication device. By calculating the correlation coefficient between the linear combination coefficients corresponding to the principal component data, it is possible to determine which principal component data of different communication devices have similarities. In order to subsequently cluster the principal component data in each communication device, the linear combination coefficients of the principal component data in different communication devices are first used for matching to obtain all matching principal components between any two communication devices.
[0055] Preferably, in one embodiment of the present invention, the matching principal component acquisition method includes:
[0056] The correlation coefficient between each principal component data of a communication device and the linear combination coefficient of all principal component data of any other communication device is calculated, and the two principal component data with the largest correlation coefficient are used as matching principal components of any two communication devices.
[0057] Preferably, in one embodiment of the present invention, the Pearson correlation coefficient is used to obtain the matching principal components of any two communication devices. The correlation coefficient calculation process is a technical means well known to those skilled in the art and will not be described in detail here. There are many methods for obtaining matching principal components. Here, only a step of obtaining matching principal components using the Pearson correlation coefficient is briefly described, and the step is as follows:
[0058] Record The first communication device The principal components are , the communication data dimension of each communication device is obtained as ,and , ,in Indicates The corresponding communication device The original dimension data.
[0059] For The first communication device principal component In terms of The first communication device principal component The correlation coefficient satisfies: With The principal component data with the largest correlation coefficient among the communication devices is , then As By reordering the matching principal components between any two communication devices, the matching relationship between all principal components of the two communication devices can be obtained. The first communication device The matching principal components are , after matching The first communication device The matching principal components are .
[0060] At this point, all matching principal components between any two communication devices are obtained.
[0061] Step S3: When using the ISODATA algorithm for clustering, all communication devices are initially clustered using preset initial parameters to obtain initial sample clusters; the difference between the matching principal components of any two communication devices is obtained; based on the difference between the corresponding variance contribution rate of any matching principal component between any two communication devices in each initial sample cluster and the matching principal component, the degree of dispersion of the initial sample cluster under any matching principal component is obtained; the cluster center of the initial sample cluster is split according to the degree of dispersion to obtain the cluster center point after the split.
[0062] When the ISODATA algorithm is used to cluster the communication data of each dimension of each communication device, initial clustering is first performed based on the distance between the corresponding principal component data of each communication device and the preset clustering center to obtain the initial sample cluster.
[0063] When the ISODATA algorithm performs clustering, the data mean of the principal component data in each communication device is usually used for clustering. However, for the principal component data that changes significantly within a preset time, only considering the mean of the principal component data for clustering will cause the principal component data that changes relatively steadily and the principal component data that changes significantly to be classified into one cluster, thereby making the clustering result inaccurate. Therefore, in the embodiment of the present invention, the difference between the matching principal components of any two communication devices is obtained based on the degree of data difference between the matching principal components and the degree of similarity of the corresponding data at each moment.
[0064] Preferably, in one embodiment of the present invention, the method for obtaining the difference includes:
[0065] The square of the difference between the first-order differences of the corresponding data at each moment under any matching principal component in any two communication devices is calculated as the first-order difference similarity, and the first-order difference similarity is negatively correlated and normalized to obtain the first-order difference similarity of the corresponding data at each moment under any matching principal component in any two communication devices; the first-order difference at each moment is the difference between the principal component data at the next moment and the principal component data at the corresponding moment; the sequence number of each moment within a preset time is normalized to obtain a weight coefficient; the first-order difference similarity of the corresponding data at each moment is weighted by the weight coefficient to obtain the similarity of the corresponding data at each moment in any matching principal component in any two communication devices within the preset time; the difference between the data means of any matching principal components in any two communication devices is used as the data difference degree; according to the data difference degree of any matching principal component in any two communication devices within the preset time and the similarity degree of the corresponding data at each moment, the difference between any matching principal components in any two communication devices is obtained.
[0066] Preferably, in one embodiment of the present invention, the method for obtaining the difference according to the similarity degree and the data difference degree includes:
[0067] The difference is obtained according to the difference calculation formula, which is as follows:
[0068] ;
[0069] In the formula, Indicates A communication device and The communication device is The differences between the matching principal components; Indicates The first communication device Matching principal components; Indicates The first communication device Matching principal components; Indicates The first communication device The data means within the matching principal components; Indicates The first communication device The data means within the matching principal components; Indicates The current moment of the matching principal component within the preset time; Indicates The time sequence number of the matching principal component within the preset time; Indicates The first communication device Matching principal components The first-order difference of the data corresponding to each moment; Indicates The first communication device Matching principal components The first-order difference of the data corresponding to each moment; represents the normalization function; Represents an exponential function with a natural constant as its base.
[0070] In the difference calculation formula, by calculating the The first communication device Matching principal components The first-order difference of the data reflects the The first communication device The matching principal components are The data changes corresponding to the moment; calculate the The first communication device Matching principal components The first-order difference of the data reflects the The first communication device The matching principal components are The data changes corresponding to the data at each moment; It indicates the A communication device and The first communication device The matching principal components are The first-order difference similarity of the data at each moment. The smaller the first-order difference similarity, the smaller the A communication device and The first communication device The greater the difference in the first-order difference similarity, the greater the difference in the first-order difference similarity. A communication device and The first communication device The difference of the matching principal components; the corresponding data credibility at different times is not the same, and the value with a larger time span than the current time is given a lower weight coefficient, so the time sequence number is used for normalization As the weight coefficient of the first-order difference similarity, the first-order difference similarity of the corresponding data at each moment is weighted to obtain the first A communication device and The first communication device The similarity of the matching principal components, The difference between the data means reflects the A communication device and The first communication device The degree of data difference between the first matching principal components; the degree of data difference is corrected by using the similarity degree to obtain the first A communication device and The first communication device The differences between the matching principal components.
[0071] It should be noted that the preset time is set to 60s, and the preset time can be set by the implementer and is not limited here.
[0072] After obtaining the difference between all matching principal components of any two communication devices, since the discreteness of each initial sample cluster after initial clustering is determined by the data mean corresponding to all communication devices in the initial sample cluster, the discreteness of the initial sample cluster corresponding to communication devices with different change trends may be smaller, resulting in inaccurate clustering results, so it is necessary to correct the discreteness of each initial sample cluster in the clustering process. Therefore, in one embodiment of the present invention, the discreteness of the initial sample cluster under any matching principal component is obtained according to the difference between the corresponding variance contribution rate of any matching principal component between any two communication devices in each initial sample cluster and the matching principal component.
[0073] Preferably, in one embodiment of the present invention, the discrete degree acquisition method includes:
[0074] The discrete degree is obtained according to the discrete degree calculation formula. The discrete degree calculation formula is as follows:
[0075] ;
[0076] In the formula, Indicates that each initial sample cluster is The degree of dispersion under the matching principal components; Indicates the number of communication devices included in the initial sample cluster; Indicates the first The first communication device The corresponding variance contribution rate of the matching principal components; Indicates the first The first communication device The corresponding variance contribution rate of the matching principal components; Indicates A communication device and The communication device is The differences between the matching principal components; Indicates A communication device and The communication device is The initial degree of dispersion between the matching principal components.
[0077] In the discrete degree calculation formula, since the variance contribution rate of the matching principal component of each communication device in the initial sample cluster is different, the first The first communication device The corresponding variance contribution rate of the first matching principal component is The first communication device The average value of the corresponding variance contribution rate of the matching principal components is taken as The weight coefficient of A communication device and The communication device is The initial discrete degrees between the matching principal components are corrected; the corrected initial discrete degrees are averaged to obtain the values of each initial sample cluster in the The degree of dispersion under the matching principal components.
[0078] After obtaining the degree of discreteness of each sample cluster under different principal components, it is determined whether each sample needs further splitting according to the given algorithm parameters. The traditional splitting method is to split the cluster center point according to the dimension with the largest variance. This method will cause some dimensions to have a large degree of discreteness but not be divided, which will cause the algorithm to need to perform multiple cycles of splitting the cluster center point before it can become stable. Therefore, in one embodiment of the present invention, the cluster center of the initial sample cluster is split based on the degree of discreteness of the initial sample cluster under all matching principal components to obtain the cluster center point after splitting.
[0079] Preferably, in one embodiment of the present invention, obtaining the cluster center point after splitting includes:
[0080] All the corresponding principal component data in each communication device in the initial sample cluster are averaged to obtain the initial cluster center; the discrete degree of all the corresponding principal component data in each communication device in each initial sample cluster is squared to obtain the cluster center adjustment vector; the deviation coefficient is preset, and the product of the deviation coefficient and the cluster center adjustment vector is used as the cluster center adjustment coefficient; the initial cluster center is split according to the cluster center adjustment coefficient.
[0081] It should be noted that the cluster center splitting step of the ISODATA algorithm is a technical means well known to those skilled in the art, and is not limited here. Only one embodiment of the present invention is briefly described. If a certain initial sample cluster needs to be further divided, the step of obtaining the cluster center after the split specifically includes:
[0082] ;
[0083] In the formula, Represents a cluster center after splitting; Represents another cluster center after splitting; represents the initial cluster center, ,in They are obtained by averaging the 1-z principal component data of all communication devices in the initial sample cluster; represents the coefficient of dispersion, which is set to 0.5; represents the cluster center adjustment vector, .
[0084] At this point, if the initial sample cluster needs to be split, the cluster center after the split is obtained.
[0085] Step S4: Obtain clustering results; perform equipment fault detection according to the clustering results.
[0086] When using the ISODATA algorithm for clustering, you need to preset the initial parameters first.
[0087] It should be noted that the setting of the initial parameters can be set by the implementer according to the specific implementation scenario, and is not limited here. Only one embodiment of the present invention is briefly described, and the initial parameters of the ISODATA algorithm are set, specifically including:
[0088] The initial number of cluster centers ; The minimum number of samples in each initial sample cluster ; The degree of dispersion of sample distribution in an initial sample cluster ; The threshold for merging initial sample clusters ; The maximum number of pairs of cluster centers that can be merged in one iteration ; Maximum number of iterations .
[0089] The sample is the communication data in each communication device; when the number of samples in the initial sample cluster is less than , then the initial sample cluster is not considered as an independent cluster; when the sample dispersion in the initial sample cluster is greater than , then the initial sample cluster is split; when the distance between the cluster centers of two initial sample clusters is less than , then the two initial sample clusters are merged.
[0090] After setting the initial parameters and improving the method for obtaining the discrete degree in the initial sample cluster, the ISODATA algorithm is used for clustering to obtain the clustering result. It should be noted that the ISODATA algorithm is a technical means well known to those skilled in the art and will not be described in detail here.
[0091] In one embodiment of the present invention, each cluster in the clustering results contains multiple communication devices. The communication devices in the cluster with the least number of communication devices are used as suspected faulty communication devices. The information load on all devices in the cluster is adjusted in time through load balancing, which can avoid information loss in time and ensure the security and stability of the communication system.
[0092] At this point, the communication equipment fault detection is completed.
[0093] In summary, the present invention obtains the communication data of each dimension of each communication device; performs dimensionality reduction on the communication data of each dimension, and removes the redundant information that may exist in the multi-dimensional communication data; uses the linear combination coefficients of the principal component data in different communication devices for matching, and obtains all the matching principal components between any two communication devices; obtains the initial sample cluster, and analyzes the discrete degree between the corresponding principal component data of each communication device in the initial sample cluster; calculates the similarity degree and data difference degree of each data between the matching principal components of any two communication devices, and then obtains the difference between the matching principal components of any two communication devices; uses the obtained difference to obtain the discrete degree of the initial sample cluster under any matching principal component, and corrects the discrete degree within the initial sample cluster; uses the discrete degree to bring into the ISODATA algorithm for clustering, obtains the clustering result, and performs equipment fault detection. The present invention performs dimensionality reduction on the communication data of each dimension of the communication device, avoids the influence of redundant information on the clustering result, and in the clustering process of the ISODATA algorithm, takes into account the influence of the change trend of the data on the discrete degree of the cluster cluster, so that the cluster center point after the split is more accurate and the clustering result is more accurate.
[0094] An embodiment of the present invention also provides a remote online monitoring system for communication equipment faults, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it can implement the steps of a remote online monitoring method for communication equipment faults.
[0095] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0096] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other. Each embodiment focuses on any two differences from other embodiments.
Claims
1. A remote online monitoring method for communication equipment failure, characterized in that: The method comprises: Obtain communication data of each dimension of each communication device; the dimensions include at least: power load, CPU utilization, memory usage, and traffic data; Performing dimensionality reduction on the multi-dimensional communication data of each communication device to obtain the principal component data of each communication device, the linear combination coefficient corresponding to the principal component data, and the variance contribution rate; obtaining the matching principal component between any two communication devices according to the correlation coefficient between the linear combination coefficients of the principal component data in different communication devices; When clustering is performed using the ISODATA algorithm, all communication devices are initially clustered using preset initial parameters to obtain initial sample clusters; the difference between the matching principal components of any two communication devices is obtained; the degree of dispersion of the initial sample cluster under any matching principal component is obtained based on the difference between the corresponding variance contribution rate of any matching principal component between any two communication devices in each of the initial sample clusters and the matching principal component; the cluster center of the initial sample cluster is split according to the degree of dispersion to obtain the cluster center point after the split; Obtaining clustering results; and performing equipment fault detection according to the clustering results.
2. A method for remote online monitoring of communication equipment failure according to claim 1, characterized in that: The matching main component acquisition method comprises: The correlation coefficient between each principal component data of a communication device and the linear combination coefficient of all principal component data of any other communication device is calculated, and the two principal component data with the largest correlation coefficient are used as matching principal components of any two communication devices.
3. A method for remote online monitoring of communication equipment failure according to claim 2, characterized in that: The correlation coefficient is the Pearson correlation coefficient.
4. A method for remote online monitoring of communication equipment failure according to claim 1, characterized in that: Differentiation methods include: Calculate the square of the difference between the first-order differences of the corresponding data at each moment under any matching principal component in any two communication devices as the first-order difference similarity, and perform negative correlation mapping and normalization processing on the first-order difference similarity to obtain the first-order difference similarity of the corresponding data at each moment under any matching principal component in any two communication devices; the first-order difference at each moment is the difference between the principal component data at the next moment and the principal component data at the corresponding moment; Normalize the sequence number of each moment within the preset time to obtain the weight coefficient; The weight coefficient is used to weight the first-order difference similarity of the data corresponding to each moment, so as to obtain the similarity of the data corresponding to each moment in any matching principal component of any two communication devices within a preset time; The difference between the data means of any matching principal components in any two communication devices is taken as the degree of data difference; The difference between any matching main components in any two communication devices is obtained according to the data difference degree of any matching main components in any two communication devices within a preset time and the similarity degree of corresponding data at each moment.
5. A method for remote online monitoring of communication equipment failure according to claim 4, characterized in that: The method for obtaining the difference according to the similarity and the data difference comprises: The difference is obtained according to the difference calculation formula, and the difference calculation formula is as follows: ; In the formula, Indicates A communication device and The communication device is The differences between the matching principal components; Indicates The first communication device Matching principal components; Indicates The first communication device Matching principal components; Indicates The first communication device The data means within the matching principal components; Indicates The first communication device The data means within the matching principal components; Indicates The current moment of the matching principal component within the preset time; Indicates The time sequence number of the matching principal component within the preset time; Indicates The first communication device Matching principal components The first-order difference of the data corresponding to each moment; Indicates The first communication device Matching principal components The first-order difference of the data corresponding to each moment; represents the normalization function; Represents an exponential function with a natural constant as its base.
6. A method for remote online monitoring of communication equipment failure according to claim 1, characterized in that: The discrete degree acquisition method comprises: The discrete degree is obtained according to the discrete degree calculation formula, and the discrete degree calculation formula is as follows: ; In the formula, Indicates that each initial sample cluster is The degree of dispersion under the matching principal components; Indicates the number of communication devices included in the initial sample cluster; Indicates the first The first communication device The corresponding variance contribution rate of the matching principal components; Indicates the first The first communication device The corresponding variance contribution rate of the matching principal components; Indicates A communication device and The communication device is The differences between the matching principal components; Indicates A communication device and The communication device is The initial degree of dispersion between the matching principal components.
7. A method for remote online monitoring of communication equipment failure according to claim 1, characterized in that: Get the cluster center points after splitting, including: Averaging each corresponding principal component to be analyzed in all communication devices in the initial sample cluster to obtain an initial cluster center; The cluster center adjustment vector is obtained by taking the square root of the discrete degree of each corresponding principal component to be analyzed in all communication devices in each of the initial sample clusters; a deviation coefficient is preset, and the product of the deviation coefficient and the cluster center adjustment vector is used as the cluster center adjustment coefficient; and the initial cluster center is split according to the cluster center adjustment coefficient.
8. A method for remote online monitoring of communication equipment failure according to claim 1, characterized in that: Performing equipment fault detection according to the clustering result includes: According to the clustering results, the class with the least number of samples among all clusters is selected as the suspected faulty communication equipment.
9. A method for remote online monitoring of communication equipment failure according to claim 1, characterized in that: The initial parameters include: The initial number of cluster centers ; The minimum number of samples in each initial sample cluster ; The degree of dispersion of sample distribution in an initial sample cluster ; The threshold for merging initial sample clusters ; The maximum number of pairs of cluster centers that can be merged in one iteration ; Maximum number of iterations ; The sample is the communication data in each of the communication devices.
10. A remote online monitoring system for communication equipment faults, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of a method for remote online monitoring of communication equipment failures as described in any one of claims 1 to 9 are implemented.
