Research On The Algorithm Of Parallel Fft And Its Applications On High Speed Rail Data Based On Cloud Computing

The progress of science and technology constantly promote the development of high speed rail and make it become an energy-saving, environmental, fast and punctual transportation vehicle and the first choice for people on travelling. To guarantee trains run safely, we need to analyze and process data collected from different parts of trains in order to grasp the running state and make some decisions.To get enough information from high speed trains, engineers’ve installed many sensors in cells of trains to collect different kinds of data like noise and vibration. However, a huge challenge is confronted when the traditional signal processing way is used for handling the increasing big data. As a new parallel processing technology, cloud computing combines different properties of grid computing, distributed computing and parallel computing. It performs very well in processing big data and managing network storage.Being a framework of cloud computing. Hadoop contains Hadoop File System (HDFS) and MapReduce—the novel parallel programming model, and is high-reliable and high-tolerant in processing. Especially, the MapReduce model applies divide-and-conquer principle that efficiently alleviates the challenge brought by the difficulty of modern programming ways when confronting big data. This thesis makes data preprocessing and FFT algorithms become parallel based on cloud computing technology. Especially, FFT is widely applied in digital signal analysis and plays an important role in the area of high speed rail data processing. As the first step, unpacking of original data based on cloud computing establishes a solid base for subsequent work like data smooth, wrong points removal and linear trend items removal, etc. Experiments for evaluating the proposed parallel unpacking algorithm show a very good performance.In order to support engineers with a systematical environment for data preprocessing, this thesis designs a high speed rail data processing system based on cloud computing that combines multiple kinds of algorithms of data preprocessing together and makes configuration authority enable for Hadoop cluster. What engineers need to do is submitting tasks to this system according to specific requirements. After that, system will transmit this task to Hadoop clusterand engineers could download the results from HDFS when finished, thus making the process of data preprocessing easier and more convenient.As a variety of discrete fourier transform (DFT). FFT is fast in algorithm processing and becomes an important tool in digital signal analysis. In addition, it is extensively used in image processing and communication technology. The same to other kinds of signal data, processing high speed rail data also demands FFT. However, traditional serial FFT algorithm could not handle large volumes of high speed rail data. Therefore, this thesis puts forward a parallel FFT algorithm based on cloud computing technology. Experiments show that our algorithm maintains same results compared with the serial algorithm and increases the efficiency to some extent. Also, it can handle large volumes of high speed rail data as demanded.