Research On Mobile Base Station Equipment Fault Early Warning System Based On Machine Learning

With the rapid development of mobile communication industry,operators have increased the construction of mobile base stations.The increase of mobile base station scale brings challenges to the daily operation and maintenance work.Although at present,the mobile base station has established DCS,SCADA and other systems to monitor the base station equipment,but these systems have problems such as data dispersion,insufficient use of historical data,low efficiency of data processing and low prediction ability.With the application of big data technology and machine learning model,the use of computer system to collect equipment operation data,integrate relevant data information,and establish machine learning methods to early warn the fault information of base station equipment has become an important direction in the field of mobile communication equipment early warning.This thesis studies how to integrate the machine learning function into the early warning system of mobile base station equipment to solve the problems of low utilization rate of historical data in the current early warning system of equipment,the early warning information needs to be judged manually,and the machine learning model cannot be used.Firstly,this thesis analyzes the requirements of the system.At present,the data of the mobile base station is generally kept in the system of the base station,the data is relatively distributed,there is no effective integration,and the historical data is not fully utilized.At the same time,the alarm information sent by the equipment sensor,such as high temperature,voltage fluctuation,etc.,needs further manual screening,and the processing efficiency is low.The early warning information of the sensor is usually sent when the equipment is about to or has problems,and the sensor monitors the abnormal data,and does not provide the prediction information about whether the equipment may fail in the future.Therefore,the core requirement of this system is to collect the fault information of each base station,integrate the data through the big data platform,and then integrate the machine learning function into the system,so that the operation and maintenance personnel can call the machine learning model through the web page for training,and predict the failure probability and type of the equipment in a certain period of time in the future.Secondly,this thesis designs the early warning system in detail.Combined with the requirements,the system needs to solve the problem of big data storage and supporting machine learning model operation through big data platform,and can call the model through the web.Therefore,this thesis designs the data collection and cleaning process of the whole system,the structure used for data storage in the big data platform,and the model training,model warning and other functions of calling machine learning model for early warning.Thirdly,this thesis introduces the implementation of each module in detail.This thesis uses kettle to realize traditional database and big data platform ETL.Through the integration of spark mlib,sklearn,PMML and Jave web,the function from front-end input to back-end big data machine learning model is realized.Then,the time series model of system configuration time window training,machine learning model and the realization of model checking are studied.Finally,using Java Web front-end to call the kettle interface,the operation and maintenance personnel can customize the data extraction function.Fourthly,this thesis tests the function and performance of the system,the module function of the system,ETL and machine learning module design cases.Finally,the performance of the big data platform is tested,and the test results meet the needs.The research in this thesis has a certain reference value for improving the processing ability of big data of mobile base station,using machine learning algorithm for equipment fault early warning,and improving the stability of base station operation.