KR100439674B1 - Apparatus and method for failure prediction of radio port using neural network - Google Patents

Abstract

The present invention provides a base station failure prediction apparatus and a method using a neural network, Management data collection unit for collecting the management data of the base station; A neural network processing unit for processing the information of the base station collected by the management data collection unit as a neural network; By including a failure possibility output unit for extracting and outputting the possibility of failure of the base station from the results processed by the neural network processing unit, after predicting a base station with a high probability of failure by using a neural network, before failure occurs in the base station By focusing on the most likely base stations, it is possible to minimize the failure of the base stations and maximize the operational efficiency.

Description

Apparatus and method for failure prediction of radio port using neural network

The present invention relates to base station failure prediction, and more particularly, to a base station failure prediction device and a method using a neural network suitable for predicting a base station with a high probability of failure using a neural network.

In general, a communication system is a system that performs information processing by combining a terminal with a remote communication line, and there are wired and wireless communication systems.

The wireless communication system is a communication system for mobile devices such as people, cars, ships, trains and aircraft, including mobile phones (mobile phones, vehicle phones), port phones, aircraft phones, mobile public phones (trains, cruise ships, Installed on express buses, radio calls, radiotelephones, satellite communications, amateur radio, and fishing service ships. These communications include the Advanced Mobile Phone Service (AMPS) system using analog methods, the CDMA and TDMA (Time Division Multiple Access) systems using digital methods, and the Frequency Division Multiple Access (FDMA) systems. , WLL (Wireless Local Loop) system.

1 is a block diagram of a general communication system.

As shown in the figure, an exchanger 1 for transmitting a call processing request of a terminal to another communication network such as a public telephone switching network through a public network or a dedicated network so that a communication service can be made; A control station (2) connected to the exchange (1) to control a base station (3); A base station (3) under the control of the control station (2), transmitting a call transfer request from the switchboard (1) to the terminal (4), and receiving a call processing request from the terminal (4); It is composed of a terminal (4) for receiving a communication service to the subscriber in the coverage area of the base station (3).

In the general communication system configured as described above, if the subscriber intends to use the communication service while having the terminal 4 within the coverage area of the base station 3, the subscriber terminal 4 information obtained through the base station 3 is exchanged. Transmit to home location register through (1). In addition, the communication system operates to perform a voice information service according to a request of the terminal 4 or to connect to another communication network to perform a communication service.

In the related art, when a failure occurs in the base station 3, measures are taken after the failure occurs.

However, this conventional technology has a problem that it is not possible to focus on the base station having a high frequency of failure because it is not known in advance which base station and in which part of the base station high frequency.

Accordingly, the present invention has been proposed to solve the conventional problems as described above, and an object of the present invention is to predict a base station with a high probability of failure using a neural network, and then, before the failure occurs in the base station, the base station with a high probability of failure The present invention provides a base station failure prediction device and a method using a neural network capable of minimizing the failure of the base station and maximizing the operation efficiency by monitoring.

Base station failure prediction apparatus using a neural network according to an embodiment of the present invention to achieve the above object,

A management data collection unit collecting management data of the base station; A neural network processing unit for processing the information of the base station collected by the management data collection unit as a neural network; The technical configuration is characterized in that it comprises a failure possibility output unit for extracting and outputting the possibility of failure of the base station from the results processed by the neural network processing unit.

Base station failure prediction method using a neural network according to an embodiment of the present invention to achieve the above object,

A first step of collecting management data of the base station; A second step of processing management data of the base station collected in the first step into a neural network; And a third step of extracting and outputting the possibility of failure of the base station from the result processed in the second step.

1 is a block diagram of a general communication system,

2 is a block diagram of a base station failure prediction apparatus using a neural network according to the present invention,

3 is a flowchart illustrating a method for predicting failure of a base station using a neural network according to the present invention;

4 is a conceptual diagram illustrating a configuration example of a neural network in FIGS. 2 and 3;

5 is a graph of a multilayer perceptron generalized to FIG. 4,

6 is a diagram showing an example of outputting a list of failure prediction upper base stations according to the embodiment of the present invention.

Explanation of symbols on the main parts of the drawings

1 switch 2 control station

3: base station 4: terminal

10: management data collection unit 20: neural network processing unit

30: Possible failure output unit

Hereinafter, an embodiment of the present invention configured as described above, a base station failure prediction apparatus using a neural network, and a method thereof will be described in detail with reference to the accompanying drawings.

2 is a block diagram of a base station failure prediction apparatus using a neural network according to the present invention.

As shown here, the management data collection unit 10 for collecting management data of the base station; A neural network processing unit 20 for processing the information of the base station collected by the management data collection unit 10 as a neural network; It is configured to include a failure probability output unit 30 for extracting and outputting a failure probability of the base station from the result processed by the neural network processing unit 20.

The management data collection unit 10, including the hardware information of the base station, the software information of the base station, the repair history information of the base station, the performance information of the base station, the failure information of the base station, the ticket issuance history information received and collected to the neural network This makes it possible to predict the possibility of failure at the base station.

In the above, the neural network processing unit 20 predicts the possibility of failure in the base station by using the performance indicator classification criteria, failure indicator classification criteria, base station capacity calculation, operator action content classification.

In the above, the neural network processing unit 20 predicts the possibility of occurrence of a failure of the base station on the basis of the number of tickets and the critical failure transfer time.

3 is a flowchart illustrating a method for predicting failure of a base station using a neural network according to the present invention.

As shown therein, a first step ST11 of collecting management data of the base station; A second step ST12 of processing management data of the base station collected in the first step into a neural network; And a third step (ST13) of extracting and outputting the possibility of failure of the base station from the result processed in the second step.

In the first step, the hardware information of the base station, the software information of the base station, the repair history information of the base station, the performance information of the base station, the failure information of the base station, the ticket issuance history information received and collected by the neural network at the base station Make it possible to predict the probability of failure.

In the second step, the probability of failure in the base station is predicted using the performance indicator classification criteria, the failure indicator classification criteria, the base station capacity calculation, and the operator action content classification.

In the second step, the possibility of failure of the base station is predicted based on the number of tickets and the critical failure time.

Referring to the accompanying drawings, the operation of the base station failure prediction device and the method using the neural network according to the present invention configured as described in detail as follows.

First, the present invention is intended to minimize the failure of the base station by using a neural network to predict the base station with a high probability of failure and to focus monitoring.

In the prior art, there was no technology for comprehensively predicting and monitoring a base station that would cause many failures. Therefore, it was impossible to predict the failure of a general base station because it was not possible to identify the association between the single element data and the occurrence of a failure.

Therefore, in the present invention, it is possible to improve accuracy in base station failure prediction by using neural network analysis using complex manpower factors (disability, performance, ticket, configuration repair history, etc.) and use it in actual operation.

So, for example, the base station-related network collection data, tickets, failover, and repair history data generated during the week of the week is aggregated by the base station and used as an input factor. The neural network analysis focuses on the list of the top base stations (for example, within the top 5%) that are more likely to fail in the vehicle.

4 is a conceptual diagram illustrating a configuration example of a neural network in FIGS. 2 and 3.

Therefore, the management input factor may have the following items.

(1) Hardware: There is a possibility of failure due to traffic increase depending on the hardware capacity and characteristics of the base station.

-Base station type: Failure factor is different for each base station type.

Frequency Allocation (FA), Number of Sectors: According to this factor, the capacity and utilization purpose of the base station are different.

-Number of main boards: Main management boards vary depending on the type and function.

(2) Software: Frequent software changes greatly affect the possibility of failure.

-Package Version: The pattern of the error that is corrected and the error that is not corrected differs according to the applied package.

-Whether PLD (Processor Load Data) is recently loaded: If PLD loading is frequent, there is a possibility of failure, and if there is a recent loading, the probability of failure is low.

(3) Manual management: The possibility of failure depends on the type of repair and recent repair.

-Recently repaired: In case of frequent repairs, the board may have a failure. In the case of recently repaired boards, the probability of failure may be high or low.

(4) Performance: Since the occurrence of a failure eventually leads to a decrease in performance, it is closely related to the failure.

Traffic: Congestion and trend of traffic, affecting the possibility of failure in terms of base station capacity.

-Traffic rate / Completion rate / Cutting rate: The trend and specificity of the communication rate / completion rate / cutting rate, and serious degradation of the communication rate, completion rate, and cutting rate can be expressed as a result of failure prediction output data.

-Fail Reason (Fail Reason): Classify and analyze hardware related call failures and call failures related to congestion. Fail listen can be analyzed and used to extract fault association.

(5) Disability: The history of past disability has a great influence on the possibility of disability in relation to the details of repair measures.

-Error occurrence time / frequency: indicates the occurrence and frequency of failure of a specific base station.

-Types of Disabilities: Analyze the importance of each type of disability.

-Elongation time: Analyze the elapsed time of fault clear.

(6) Tickets: Ticket issuance information is closely related to major failures.

-Ticket type: distinguish between automatic ticket generated by the system and manual ticket generated by the operator.

-Leave time: The ticket making time is processed by the frequency of a certain time (for example, 10 minutes).

Management information input by such a management input factor is analyzed through a neural network.

Therefore, the range of each input factor is simplified for the purpose and purpose of analysis.

-Base station capacity: Manage by appropriate capacity, insufficient capacity and spare capacity according to type.

-Software version: Manage as update according to failure, update to improve performance, etc.

-Recently repaired or repaired: repair due to failure, repair to improve performance, etc.

-Performance: Control up, down and down on the basis of appropriate management value.

-Disability: Manage the number of occurrences and average time of disability according to the type of disability as upper, middle and lower according to appropriate amount.

FIG. 5 is a graph of a multilayer perceptron generalized to FIG. 4. FIG.

Therefore, the neural network has a property of a network, as shown in FIGS. 4 and 5, which is a graph composed of edges having weights and directions. Each node is modeled as a neuron and has an activation level and an output. The activity value represents the extent to which the cell is "excited" and the output value is determined nonlinearly in proportion to the degree of excitation. Each edge is called a synapse and is actually weighted by modeling the synapse. The structure of the network depends on how each node is connected to each other. Here, learning becomes a problem of how the weight of synapse is determined by external stimulus (input / output).

One node receives the output values of other nodes in proportion to the weight of the corresponding synapse. Therefore, the active value of a node is calculated internally of an input vector composed of outputs of other nodes and a weight vector composed of weights of corresponding synapses. This value is converted into an activation function that is a nonlinear function and output.

In addition, multilayer perceptron is a layered neural network composed of input layer, hidden layer, and output layer. Input x entering the input layer is converted from the hidden layer to z and from the output layer to y. So if we graph the linear regression model, there are no z1 and z2, and instead all x nodes in the input layer are directly connected to y. In this case, the weight of the synapse connecting the x and y nodes is equal to the coefficient in the linear regression. Thus, it can be seen that the multilayer perceptron includes a linear regression model.

The base station information analysis through the neural network targets the extraction of the possibility of the occurrence of a borrower failure of the base station.

In order to create a neural network model using input data, a value corresponding to a target may be defined as a more detailed field.

In determining the target, the presence of a critical disorder can be considered. However, the target does not necessarily coincide with the presence of a critical failure, and an appropriate neural network model can be obtained when the number of tickets and the critical failure time are met.

In addition, the base station failure prediction model needs to be periodically updated with an appropriate prediction model according to the change of network conditions. To this end, the accuracy of the prediction results should be evaluated over a certain interval (for example, weekly), and if the accuracy decreases, the prediction model should be updated.

FIG. 6 is a diagram showing an example of outputting a list of failure prediction upper base stations according to an embodiment of the present invention, and shows the final results by this predictive model.

Thus, the horizontal axis of FIG. 6 is base station information, and has a unique value for each target base station. In the graph of FIG. 6, the probability of failure is derived by analyzing the management input factor including the number of major failure predictions and the predicted failure time by a neural network. Therefore, it can be seen that the more base stations with high probability of failure, the more failures actually occur.

As described above, the present invention focuses on a base station with a high probability of failure before predicting a base station with a high probability of failure using a neural network and before a failure occurs at the base station.

Although the preferred embodiment of the present invention has been described above, the present invention may use various changes, modifications, and equivalents. It is clear that the present invention can be applied in the same manner by appropriately modifying the above embodiments. Accordingly, the above description does not limit the scope of the invention as defined by the limitations of the following claims.

As described above, the base station failure prediction apparatus and the method using the neural network according to the present invention after predicting a base station with a high probability of failure using the neural network, and focuses on the base station with a high probability of failure before the failure occurs in the base station By doing so, it is possible to minimize the failure of the base station and maximize the operational efficiency.

Claims (8)

delete A management data collection unit for collecting management data including hardware information, software information, repair history information, performance information, failure history information, and ticket issuance history information of a base station associated with a failure from each base station; A neural network processing unit for classifying the elements of the collected management data according to the contents thereof, and then applying them as input factors to a neural network of a neuron model composed of weights and directional edges to perform neural network analysis; And a failure occurrence output unit configured to extract processing results of the neural network processing unit as a possibility of failure of a corresponding base station, and output a failure possibility of each base station. According to claim 2, The neural network processing unit, A base station failure prediction device using a neural network, characterized by predicting the occurrence of failure in the base station using the performance indicator classification criteria, failure indicator classification criteria, base station capacity calculation, operator action content classification. According to claim 2, The neural network processing unit, A base station failure prediction device using a neural network, characterized in that it is configured to predict the probability of a failure of the base station on the basis of the number of tickets and the critical failure time. delete A first step of collecting, from each base station, management data including hardware information, software information, repair history information, performance information, failure history information, and ticket issuance history information of the base station associated with a failure; A second step of classifying each element of the collected management data according to its contents, and then applying them as input factors to a neural network of a neuron model composed of weights and directional edges to perform neural network analysis; And extracting the neural network analysis result as a possibility of failure of the base station and outputting a possibility of failure of each base station. The method of claim 6, wherein the second step, A method for predicting failure of a base station using a neural network, comprising performing performance indicator classification criteria, failure indicator classification criteria, base station capacity calculation and operator action classification. The method of claim 6, wherein the second step, A method for predicting failure of a base station using a neural network, characterized in that it is performed to predict a probability of occurrence of a failure of a base station under the number of tickets and the critical failure propagation time.