Design Of On-borad Track State Monitor System

Since railway-trains have been raised speed continuously, the healthy condition of trains becomes a key factor to ensure its safeness, satisfaction and reliability. The conventional way to maintain is mainly used to do regular checking with Track Inspection Car, examine and repair the fault according the checking result afterwards.As the age of information technology and Big Data coming, the OB-TSM System has been put forward in this issue which will be installed in the steering room. We are access to an all around evalution about the condition of the rail surface whether it is flat or smooth through the vibration acceleration in horizontal and vertical directions during operational process of track car with embedded pattern recognition algorithm.This paper focuses on three sections involving hardware and software of embedded system and diagnosis algorithm. This system hardware support is to s5pv210 processor with arm Cortex TM-A8 kernel as a system of the central processing unit, the accelerometer MMA7455 as a data acquisition sensor, and other modules used the development board OK210’s factory design. Linux is the master operating system. Under this system, we have designed device driver, fault diagnosis program, fault evaluation program by Linux Application and Qt GUI。In addition, diagnosis algorithm is accomplished through C/C++ in Linux. Firstly, we extracted time domain signal characteristic vectors and using the improved BP neural network to classify the feature vectors.Then, we introduced the method of fault evaluation, it calculated the probability whether it will break down by synthesizing several testing points as its neural network.In the end, it comes to an experimental scheme in this paper. The scheme aims at specific neural network built by specific track car,which will solidify the structured neural network and integrate it in the system software. Because of the limited condition, we can only test two kinds of data that is smooth or not. The consequence of the experiment shows that self-designed improved BP neural network have some capacity of recognition, and all testing sample passed though the fault evaluation program. It shows this system plays a significant role in engineering application. Compared with traditional track inspection method, this system simplifies the checking work and it’s economical, easy to carry as well as reliable. It’s also an indication that it tends to develop to direction of production, information and intelligence in engineering field.