At present, the paces of ICT industry fusion and the transformation of telecom operators’ cloud reconstruction are gradually accelerating. The scale and complexity of telecommunication system are increasing day by day. The operators are facing increasing pressure and challenge in network operation and maintenance. There are great gap between the traditional operation and maintenance mode and the advanced network itself, the difficulties and challenges in network maintenance has become increasingly prominent.The deep learning ability of artificial intelligence can extract the implicit correlation features in the massive operation and maintenance data, trace the original fault, achieve the fault analysis through summarizing the fault by the common features.
The whole process is divided into two stages: knowledge generation and application.
●Fault acquisition and input: acquisition of current network fault data as the training data sources of the diagnostic knowledge base.
●Fault feature crawling: through data analysis and pretreatment of the original fault data from current network, extract the fault performance characteristics.
●Training and knowledge base output: through training, analysis of fault generation and disappearance of the associated information for the processed data using artificial intelligence algorithm , get the knowledge base of the fault characteristics.
Knowledge Base Application:
●Network status monitoring: real-time monitoring and alarm, traffic, packet loss, operational log and other information.
●Fault matching and diagnosis: match with the current network monitoring data, and after the successful matching, diagnose each one from big to small according to the probability .
●Decision-making: confirm the existence of the fault, give the fault point and processing recommendations if the fault do exist.
●Fault repair: trigger a failure automatically recovers or send request for repair.
Through the automatic generation of knowledge base and strategy, predictable operation and maintenance, ZTE uSmartInsight can achieve:
● Intelligent prevention before event: intelligent health testing, alarm prediction, intelligent reminder.
● Intelligent processing in event: alarm suppression, RCA analysis, associated alarm filtering, intelligent scheduling.
● Intelligent learning after event: update the prediction knowledge base, form a new features and new rules by self-learning.
Typical application scenario
● Fault analysis: comprehensive analysis and identify the cause of the problem for failure on matter in same network element or in different network elements, no matter in same professional network or in different professional network
● Abnormal prediction: Through the prediction algorithm and historical data analysis, we can predict the value in next step during the middle and short cycle range of the network health degree, and then judge whether the forecast value will exceed the threshold to decide whether the fault will occur in the future.
● Wireless capacity adaptation: According to the CU changes, learn and predict the future trend of CU, automatically adjust the cell capacity parameters, achieve a dynamic balance of user access capacity and user experience.
● Routing optimization: dynamic optimization of routing strategies and scheduling resources, improve the efficiency of transmission and network utility; achieve bandwidth adjustment, routing adjustment, priority protection and other functions; high process efficiency, automated, meet the unexpected demand.
ZTE uSmartInsight provides intelligent maintenance capability of fault handling for operators, thus reduces the manpower cost of maintenance, improves the accuracy of fault handling, improves efficiency of resource utilization, positions faults fast and solves quickly, reduces the complaint rate and improves the user experience.