Storage And Analysis Of Vehicle Road Test Data Based On Cloud Computing

Vehicle manufacturers organize and carry out various and rigorous whole vehicle test in the later stage of development,the vehicle road tests play an important role among them.They will generate massive CAN(Controller Area Network)bus full channel data and a large number of fault information.However,because of the huge size of the data,many problems occur when the vehicle manufacturers use them.These problems include: 1)The costs of the storage rise fast.2)The reports generated from the big data are becoming more and more difficulty.3)The abnormal driving behavior can't be found automatically from the test data.Against this background,this paper develops a platform to manage and analyze the road test data through using Hadoop and Spark framework,realizing data gathering,data storage,data analysis and report presentation.Firstly,this paper analyzes the requirement of the platform from the aspects of function and non-function,and creates a system use case model.Then this paper designs the architecture from the aspects of technical,logic and deployment.This platform is broken down into four subsystem,including data gathering subsystem,data storage subsystem,data analysis subsystem and report presentation subsystem.Data gathering subsystem is responsible for gathering of CAN bus data and fault information,which is temporarily saved in flash card in vehicle recorder and controller and will be transferred to this platform.Data storage subsystem is responsible for storing these original test data files and importing into HBase after being decoded.Data analysis subsystem is responsible for the analysis of CAN bus data and fault information.It can also mine driving to find abnormal driving behaviors.Report presentation subsystem will present the result of analysis on web.Secondly,this paper is focus on the following key technologies: 1)The original test data files are stored in HDFS(Hadoop Distributed File System)and they can be visited like local.Using HBase to store test data instead of centralized file storage can combine with Hive data warehouse and Spark analysis framework.2)K-means algorithm is applied to find abnormal test data.3)Hive and SparkSQL framework are used to analyze the data in the HBase.4)Navy-bayes algorithm is applied to mine road test data to ultimately find the abnormal driving behaviors.At last,this paper introduces the test environment and describes how to perform the functional testing,non-function testing and fault tolerance testing respectively.Functional test cases are designed and executed based on user requirements.Performance test is carried out from the aspects of query response time,the speed of data importing into HBase,the relationship between the number of HBase cluster nodes and the performance.A fault simulate method is used to perform fault-tolerant testing.The test results show that the platform has achieved the expected goal.