| The rail AG system as one of the main equipment of rail transit, performancerequirements are also increasing. There are several insufficiencies in the AG systemdevelopment process: access test data mainly depends on the actual human testing, need to doa lot of experiments and there are some special circumstances in experimental way are notdetected. This traditional way of research and development will lead to the determination ofthe access algorithm to cost more time, can’t meet the demand of fast and accurate, while alsoincreasing the cost of research and development.The two-way scissor AG as the research object, build a simulation model in3DS MAX,using virtual reality technology, simulates the work of a real device in Virtools, than extractedfrom human database with a variety of body characteristics, using virtual character movementto solve the problem of a large number of experiments. Comparing with several commonlyused pattern recognition method determine BP neural network as the theoretical basis of railtransit AG prevailing strategy. Extracted from the Virtool prevailing situation, the sensor statevalues as the target input values, the access channel status value as the target output values.Training BP neural network, then use the sensor data to test the trained BP neural network anddetermine the appropriate rail transit AG BP neural network access policy. The Stateflowcomplete logic flow chart of the main functions of the AG, use the automatic code generationfeature to automatically generate code, modified structures in the Microsoft Visual Studio.NET2003compiler environment the BB module functions, implement a custom BB modulestructures, and reality the whole brake system.Simulation and test results show that: AG realize normal passage detection, intrusiondetection and trailing detection function in Virtools; trained BP neural network can be used asthe prevailing strategy of the rail transit AG; Using Stateflow generated code redundancy isless, portable use after revising. |
PDF Full Text Request