TWI700907B - Diagnosis method for optical fiber loop obstacle - Google Patents

Abstract

The present invention is a diagnosis method for an optical fiber loop obstacle using artificial intelligence technology, which is characterized in that the associated network management parameters are aggregated, the feature extraction of the associated network management parameters is performed, and implementation and analysis are applied assisted by the artificial intelligence technology to obtain the obstacle root cause prediction model. Through the application of the obstacle root cause prediction model, the cause of the fiber-optic broadband loop obstacle can be accurately determined, so that a large amount of human and material resources can be saved.

Description

Optical fiber loop failure diagnosis method

The present invention relates to a network obstacle detection technology, especially a method for diagnosing optical fiber loop obstacles.

Different from the traditional point-to-point extension architecture (Point to Point; P2P), it uses Point to Multi-Point (P2MP) and Fiber To The Home (FTTH) deployment The architecture is based on the optical line terminal equipment (Optical Line Terminal; OLT) through the optical splitter (Optical Splitter) splitting and transmitting to multiple optical network units (Optical Network Unit; ONU), that is, the client equipment, this method can greatly reduce the optical fiber The usage of optical distribution network (Optical Distribution Network, ODN) is reduced, but the optical splitter of the physical layer optical fiber network has a multi-stage structure, which makes the composition of ODN quite complicated, resulting in optical fiber obstacle inspection operations Quite difficult. In the prior art, some people have proposed a monitoring method using an optical time domain reflectometer (OTDR). This method takes about 1 to 2 minutes for the optical time domain reflector to test each route, and the equipment is quite expensive. , With the large-scale adoption of fiber-to-the-home networks, this test method will take a long time to test one round, so the timeliness of the mastery shows that A considerable shortcoming, and a substantial increase in capital expenditures (Capital Expenditure or CAPEX). It can be seen from the above that there are still many deficiencies in the above-mentioned existing obstacle detection methods, which need to be improved.

In summary, if a network obstacle detection technology can be found, in particular, through the aggregation and integration of multiple parameters and automatic prediction of the cause of obstacles, so as to provide maintenance personnel with maintenance suggestions, it should effectively improve the efficiency of inspection and repair and save costs. It will become a goal that people in the technical field eagerly pursue.

The purpose of the present invention is to propose a network obstacle detection technology, which gathers relevant network management parameters for comprehensive judgment for feature extraction, and then applies artificial intelligence (AI) technology-assisted implementation methods and analysis to achieve accurate judgments on optical fiber broadband The cause of circuit obstacles, so as to achieve the goal of saving human and material resources.

In order to achieve the above and other objectives, the present invention provides a method for diagnosing optical fiber loop failure, which includes: providing network management information of the optical transceiver; after performing data cleaning of the network management information, the cleaned network management information is based on the fiber to the home obstacle Features clustering and classification and feature extraction; the network management information that has been feature extraction is used to label the root cause of obstacles; and the rules of artificial intelligence are used to model the root cause of obstacles to generate a prediction model of the root cause of obstacles, where the root cause prediction model is in the circuit to be tested During actual measurement, the real-time circuit inspection data of the circuit to be tested is input after feature extraction, so as to obtain the judgment result of the root cause of the obstacle of the circuit to be tested.

In the aforementioned method, the step of providing the network management information of the optical transceiver refers to obtaining the network management information from the passive optical network (xPON) network management system.

In the aforementioned method, the optical fiber to the home obstacle characteristics include the characteristic fields of ONU LOS status, optical distribution network (ODN) status, error quality status, laser operation status, and device connection status .

In an implementation aspect, the events included in the optical network unit signal loss state include: normal, LOS event>0, PON disconnection rate>50%, and unjudgeable.

In an implementation aspect, the events included in the state of the optical distribution network include: normal, fiber bending obstacles, fiber connector obstacles, low uplink and downlink optical loss, and unjudgeable.

In an implementation aspect, the events included in the error quality status include: normal, uplink error barrier, downlink error barrier, and inability to judge.

In an implementation aspect, the events included in the laser operating state include: normal, abnormal operation of optical line terminal equipment/optical network unit (OLT/ONU), optical line terminal equipment/optical network unit (OLT/ONU) ) The received optical power is too low, the temperature of the optical line terminal equipment/optical network unit (OLT/ONU) is too high, and the optical layer information of the passive network unit (ONU) cannot be judged.

In an implementation aspect, the events included in the device connection status include: normal, optical network unit (ONU) disconnection, optical network unit (ONU) power off, optical network unit (ONU) registration failure, Optical line terminal equipment (OLT) port service abnormalities, other abnormalities, and inability to judge.

In the foregoing method, the root cause of the obstacle includes: fiber connector loss obstacle, fiber connector reflection obstacle, fiber bending obstacle, optical line terminal equipment/optical network unit (OLT/ONU) overheating obstacle, optical line terminal equipment/optical network Optical Module Obstacles of Optical Line Unit (OLT/ONU), Optical Line Terminal Equipment/Optical Network Unit (OLT/ONU) Receiving Module Obstacle, Optical Network Unit (ONU) Power Off, Rogue ONU Obstacle, User Spindle Optical cable disconnection, optical network list The unit (ONU) is abnormally disconnected, the optical network unit (ONU) is unsuccessfully registered, the optical network unit (ONU) has been interrupted within a week, and the optical network unit management control interface (OMCI) is abnormal.

In the aforementioned method, the steps of applying the rules of artificial intelligence to modeling the root cause of obstacles refer to the use of Naïve Bayesian, C4.5 Decision Tree, Support Vector Machine, and Random Forest (Random Forest) machine learning (Machine Learning) and multi-layer perceptron (MultiLayer Perceptron) artificial neural network artificial intelligence method for modeling, and with accuracy (Accuracy), precision (Precision) and recall (Recall) Evaluate the established predictive model for the root cause of the obstacle.

In summary, the optical fiber loop obstacle diagnosis method proposed in the present invention uses the feature extraction of the relevant networked tube parameters to aggregate and comprehensively judge, and the field experts with optical fiber and professional experience in repairing and repairing are responsible for marking the root cause of the obstacle. Using artificial intelligence technology for modeling, the obtained obstacle root cause prediction model can provide a diagnosis of fiber broadband circuit obstacles. From the above, we can use the labeling operation of the root cause of the obstacle, and then the classification results of Feature Extraction can be used to Machine learning and artificial intelligence methods are used to model the root cause of obstacles, and the accuracy, precision and recall rate are used to evaluate the pros and cons of the root cause prediction model. This model can be used for the circuit to be tested for diagnosis of the root cause of the obstacle. According to the present invention The method is used for modeling, so it can accurately determine the cause of fiber broadband loop obstacles, so as to save manpower and material resources. In practical applications, it can reduce the capital expenditure (CAPEX) and operating expenses (Operating Expenditure, OPEX).

1. Loss of optical network unit signal status

2‧‧‧Optical distribution network status

3‧‧‧Error quality status

4‧‧‧Laser operating status

5‧‧‧Device connection status

6‧‧‧Optical line terminal equipment

7‧‧‧Optical Network Unit

8‧‧‧Optical Distribution Network

9‧‧‧Circuit history data

10‧‧‧Feature extraction

11‧‧‧Classification of obstacles

12‧‧‧Artificial Intelligence Modeling

13‧‧‧Model Evaluation

14‧‧‧Real-time circuit inspection data

15‧‧‧Feature extraction

16‧‧‧Obstacle root cause prediction

17‧‧‧Classification of obstacles

S11-S14‧‧‧Step

Please refer to the following detailed description of the present invention and its accompanying drawings to further understand the technical content of the present invention and its objectives and effects. The accompanying drawings are described as follows.

Figure 1 is a step diagram of the method for diagnosing an optical fiber circuit failure of the present invention.

Figure 2 is the network management parameter aggregation feature extraction classification map.

Figure 3 shows the prediction of the root cause of FTTH obstacles.

The following describes the technical content of the present invention with specific specific embodiments. Those familiar with the art can easily understand the advantages and effects of the present invention from the content disclosed in this specification. However, the present invention can also be implemented or applied by other different specific embodiments. In order to facilitate the reviewers to understand the technical features, content and advantages of the present invention and its achievable effects, the present invention is combined with the drawings and described in detail in the form of embodiments as follows. The drawings used therein are: The subject matter is only for the purpose of illustration and auxiliary description, and may not be the true proportions and precise configuration after the implementation of the invention. Therefore, it should not be interpreted in terms of the proportions and configuration relationships of the accompanying drawings, and should not limit the scope of rights of the invention in actual implementation. Hexian stated.

Figure 1 is a step diagram of the method for diagnosing an optical fiber circuit failure of the present invention. As shown in the figure, in step S11, the network management information of the optical transceiver is provided. The main purpose of this step is to obtain the network management information of the optical transceiver. Specifically, the original information of the network management can be read by the passive optical network (xPON) network management system.

In step S12, after performing data cleaning of the network management information, the cleaned network management information is collected, classified and feature extracted according to the obstacle characteristics of the fiber to the home. Specifically, obtain the network management information in the previous step, in order to avoid data format inconsistencies, missing values and invalid values For abnormal conditions, it will first go through Data Cleaning, and the cleaned network management information will be feature extracted to facilitate subsequent analysis. The feature extraction refers to the fiber to the home (FTTH) obstacle feature with a total of five features The fields are the optical network unit signal loss (ONU LOS) status, the optical distribution network (ODN) status, the error quality status, the laser operation status, and the device connection status. The present invention is based on these five feature columns. The feature extraction method is to collect all input field data including numeric and categorical data, gather them according to the correlation of field attributes, and then classify them by experts or empirical rules. This feature extraction technology can increase The predicted success rate of obstacle root cause modeling.

In step S13, the root of the obstacle is marked on the network management information that has been feature extracted. In short, this step is to mark the root cause of the obstacle in the extracted network management information. The root cause of the obstacle can be divided into sixteen items, including: fiber connector loss obstacle, fiber connector reflection obstacle, fiber bending obstacle, light Road terminal equipment/optical network unit (OLT/ONU) overheating obstacles, optical line terminal equipment/optical network unit (OLT/ONU) optical module obstacles, optical line terminal equipment/optical network unit (OLT/ONU) Receiving module failure, optical network unit (ONU) power failure, rogue ONU failure, user spindle optical cable disconnection, optical network unit (ONU) abnormal disconnection, optical network unit (ONU) unsuccessful registration, The optical network unit (ONU) was interrupted within a week and the optical network unit management control interface (OMCI) was abnormal.

In step S14, the rule of artificial intelligence is used to model the root cause of the obstacle to generate the root cause prediction model of the obstacle, wherein the root cause prediction model of the obstacle is used for the real-time circuit inspection data of the circuit to be tested when the circuit to be tested is tested. Input after extraction, so as to obtain the judgment result of the root cause of the circuit under test. In this step, the processed network management information will be modeled by artificial intelligence technology, that is, it will cover five characteristic fields and sixteen obstacle sources marked The data applies the laws of artificial intelligence to model the root causes of obstacles.

In a specific embodiment, the so-called modeling of the root of obstacles refers to the use of Naïve Bayesian, C4.5 decision tree, Support Vector Machine, and Random Forest The artificial intelligence method of machine learning and MultiLayer Perceptron (MultiLayer Perceptron) artificial neural network is used for modeling, and after the modeling, the accuracy, precision and recall are evaluated by the accuracy rate (Precision) and recall rate (Recall). The pros and cons of the established obstacle root cause prediction model.

To sum up, the present invention aims at the original network management information using the fiber to the home (FTTH) service. After data cleaning, the data is cleaned to avoid abnormal conditions such as inconsistent data format, missing values, and invalid values. The cleaned network management information first executes the feature extraction process. Then, domain experts with professional experience in repairing professional experience or through empirical rules will classify and label the network management information. Among them, the aforementioned feature extraction method is to include all input field data including numerical and categorical data, and aggregate according to the correlation of field attributes It is then classified by experts or empirical rules, and the prediction success rate of the obstacle root cause modeling can be increased through feature extraction technology. After that, the classification results of the clustered feature extraction are respectively used in naive Bayes, C4.5 decision trees, support vector machines, Machine learning such as random forest and artificial intelligence methods of multi-layer perceptron artificial neural network model the root cause of obstacles, and finally evaluate the pros and cons of the model with accuracy, precision and recall.

It can be seen from the above that numerous obstacle factors in the fiber-to-the-home circuit make the inspection work cumbersome and difficult. Experts with long-term experience in this field are required to be competent and quickly complete the obstacle removal work. In order to reduce the professional threshold and complicated process of inspection and repair, the present invention provides rapid prediction of the root cause of fiber to home obstacles. Its operation is simple and economical. It mainly extracts features of similar network management parameters first, and then uses AI technology. Anticipate the causes of obstacles and provide maintenance personnel with recommended maintenance criteria Yes, it can not only save time for repeated inspections and improve efficiency, but also greatly improve user satisfaction.

Figure 2 is the network management parameter aggregation feature extraction classification map. As shown in the figure, the figure shows the network management parameter aggregation classification feature extraction of the present invention. The feature extraction of the comprehensive diagnosis and classification of the collection related network management parameters is mainly divided into the optical network unit signal loss (ONU LOS (Loss Of Signal; LOS)) status 1, optical distribution network (ODN) status 2, error quality status 3, laser operation status 4, and device connection status 5, including five characteristic fields. Among them, the ONU LOS status 1 The item can be obtained by the optical network unit (ONU) 7, the item of the ODN status 2 can be obtained by the optical distribution network (ODN) 8, the error quality status 3, the laser operation status 4, and the equipment connection status 5 The items can be obtained by optical line terminal equipment (OLT) 6 and optical network unit (ONU) 7 respectively.

Regarding network management parameter aggregation and classification feature extraction, there are five feature fields according to the characteristics of fiber to home obstacles, which are described as follows. ONU LOS status 1 includes events such as ONU operating normally, LOS events> 0, PON disconnection rate> 50%, and cannot be judged; ODN status 2 includes events such as ODN operating normally, fiber bending obstacles, fiber connector obstacles, upstream and downstream light Loss is too low and cannot be judged; Error quality status 3 includes events such as error quality status normal, uplink error barrier, downlink error barrier and unjudgeable; laser operation status 4 includes events such as normal laser operation, OLT /ONU operation is abnormal, OLT/ONU received optical power is too low, OLT/ONU temperature is too high, there is no ONU optical layer information and cannot be judged; equipment connection status 5 includes events such as equipment connection normal, ONU disconnection, ONU disconnection Electricity, ONU registration failure, OLT Port service abnormality, other abnormalities and inability to judge.

Figure 3 shows the prediction of the root cause of FTTH obstacles. As shown in the figure, it mainly explains FTTH The establishment and subsequent application of the obstruction root cause prediction model. In the establishment of the obstruction root cause model, firstly, the circuit history data concerning the network management parameters are retrieved, such as laser optical layer data, circuit quality data, equipment alarm data, ONU registration data and ONU operation Status, etc., after extracting the relevant network management parameters by feature 10, domain experts with optical fiber and professional experience in repairing and repairing are responsible for the obstacle classification and labeling 11 of the root of the obstacle, and then the data after the obstacle classification and labeling 11 is applied to the rules of artificial intelligence Carrying out artificial intelligence modeling 12, the present invention provides sixteen obstacle root cause diagnosis results. The obstacle root cause classification items include fiber connector loss obstacle, fiber connector reflection obstacle, fiber bending obstacle, OLT/ONU temperature overheating obstacle, OLT/ONU optical mode Group obstacles, OLT/ONU receiving module obstacles, ONU power failure, Rogue ONU obstacles, user spindle fiber optic cable disconnection, ONU abnormal disconnection, ONU registration unsuccessful, ONU transient disconnection within a week, and optical network The unit management control interface (ONU Management and Control Interface; OMCI) is abnormal. Finally, the model is evaluated with accuracy, precision, and recall rate 13 to judge the quality of the model.

In addition, for the application of the obstacle root cause model, real-time circuit inspection data 14 under the actual test of the circuit to be tested is obtained. Similarly, the process of feature extraction 15 is also required to input the data completed by feature extraction 15 into the above-mentioned established obstacle The root cause prediction model is used to predict the root cause of the obstacle 16, so that the result of the predicted obstacle classification 17 of the FTTH root cause of the obstacle can be obtained.

For example, when the ONU LOS status is that the ONU is operating normally, the ODN status is the fiber bend barrier, the error quality status is the downstream error barrier, the laser operation status is the ONU receiving optical power is too low, and the device connection status is the device connection. Under normal conditions, input the above data into the obstacle root cause prediction model that has been modeled using random forest (the model has been evaluated through accuracy, precision and recall criteria), then the obstacle root cause prediction model can determine that the root cause of the obstacle is fiber bending obstacle.

Compared with the prior art, the method for diagnosing optical fiber loop obstacles provided by the present invention has the following advantages: (1) After the present invention uses relevant network management parameters to gather feature extraction, it is necessary to seek field experts with optical fiber and professional experience in repairing Participate in the task of labeling the root causes of obstacles, and then apply the classification results of feature extraction to model the root causes of obstacles using the rules of artificial intelligence, and then obtain the predicted obstacle classification results of the root of FTTH obstacles. The technology can quickly diagnose the abnormality of optical fiber circuit obstacles. Adding value through the network management information can provide a feasible, reliable, simple and economical passive optical fiber network obstacle diagnosis method; (2) The present invention can find the predicted obstacles of the root cause of FTTH obstacles on the access network ODN, OLT and ONU , To achieve the goal of rapid and large-scale diagnosis; (3) The present invention reduces the professional threshold and cumbersome and difficult inspection and repair, and the introduction of AI technology can play an auxiliary effect and provide better service quality; (4) The present invention introduces AI Technology is used to diagnose obstacles in broadband networks. Through various algorithms based on big data collection, the system can provide maintenance personnel with recommended maintenance targets after intelligent analysis. This not only saves time for repeated inspections and improves efficiency, but also improves efficiency. On the other hand, it can also greatly improve user satisfaction; and (5) The present invention can reduce network maintenance personnel costs, and can ensure the reliability and stability of passive optical networks, thereby improving maintenance efficiency and its economic benefits very obvious.

The above detailed description is a specific description of a feasible embodiment of the present invention, but this embodiment is not intended to limit the patent scope of the present invention. Any equivalent implementation or modification without departing from the technical spirit of the present invention shall be included in The scope of patent application in this case.

S11-S14‧‧‧Step

Claims (9)

A method for diagnosing optical fiber loop obstacles, which includes: providing network management information of the optical transceiver; after performing data cleaning of the network management information, all input field data of the network management information containing numerical and categorical data are based on the optical fiber Home-to-home obstacle characteristics, which collect and classify the correlation of field attributes to perform feature extraction. The fiber-to-home obstacle characteristics include ONU LOS status, optical distribution network (ODN) status, and error codes. The feature fields of quality status, laser operating status, and equipment connection status; mark the root cause of obstacles with the extracted network management information; and apply the rules of artificial intelligence to model the root cause of obstacles to generate a predictive model of the root cause of obstacles. Among them, When the circuit under test is actually tested, the obstacle root cause prediction model is used for inputting real-time circuit inspection data of the circuit under test after feature extraction, so as to obtain the judgment result of the obstacle root of the circuit under test. The method for diagnosing optical fiber loop obstacles as described in the first item of the scope of patent application, wherein the step of providing the network management information of the optical transceiver refers to obtaining the network management information from the passive optical fiber network (xPON) network management system. For the optical fiber loop failure diagnosis method described in the first item of the patent application, the events included in the signal loss state of the optical network unit include: normal, LOS event>0, PON disconnection rate>50%, and unjudgeable. According to the method for diagnosing optical fiber loop obstacles described in item 1 of the scope of patent application, the events included in the state of the optical distribution network include: normal, optical fiber bending obstacles, optical fiber connector obstacles, upstream and downstream optical loss is too low, and cannot be judged. For the optical fiber loop failure diagnosis method described in the first item of the scope of patent application, the events included in the error quality status include: normal, uplink error failure, downlink error failure, and inability to judge. For the optical fiber loop fault diagnosis method described in the first item of the patent application, the events included in the laser operating state include: normal, abnormal operation of optical line terminal equipment/optical network unit (OLT/ONU), optical line Terminal equipment/optical network unit (OLT/ONU) receiving optical power is too low, optical line terminal equipment/optical network unit (OLT/ONU) temperature is too high, passive network unit (ONU) optical layer information and unable to judge. The optical fiber loop failure diagnosis method described in the first item of the scope of patent application, wherein the events included in the connection status of the equipment include: normal, optical network unit (ONU) disconnection, optical network unit (ONU) power failure , Optical network unit (ONU) registration failure, optical line terminal equipment (OLT) port service abnormality, other abnormalities, and inability to judge. For example, the optical fiber circuit obstacle diagnosis method described in the first item of the patent application, wherein the root cause of the obstacle includes: fiber connector loss obstacle, fiber connector reflection obstacle, fiber bending obstacle, optical line terminal equipment/optical network unit (OLT/ ONU) overheating barrier, optical line terminal equipment/optical network unit (OLT/ONU) optical module barrier, optical line terminal equipment/optical network unit (OLT/ONU) receiving module barrier, optical network unit (ONU) ) Power failure, Rogue ONU obstruction, user spindle fiber optic cable disconnection, abnormal disconnection of optical network unit (ONU), unsuccessful registration of optical network unit (ONU), optical network unit (ONU) within a week It is disconnected and the optical network unit management control interface (OMCI) is abnormal. For example, the optical fiber circuit obstacle diagnosis method described in the first item of the scope of patent application, wherein the step of applying artificial intelligence to model the root cause of the obstacle refers to the use of naive Bayesian (Naïve Bayesian), C4.5 Decision tree (Decision tree), Support Vector Machine (Support Vector Machine), Random Forest (Random Forest) machine learning and MultiLayer Perceptron (MultiLayer Perceptron) artificial neural network artificial intelligence methods Modeling and evaluating the established root cause prediction model of the obstacle with accuracy, precision and recall.

Images