Research On Embedded Computing Technology In The Three-way Data Array Analysis Of Complex Samples

Three-dimensional Fluorescence Spectrometry is a new fluorescence analysis technology for complex chemical system. It has the important application value in the life science, environment, medicine and petrochemical industry. At present, this analytic methodology already has a mature development both on theory and experiment. But because of the lack of dedicated computing device, there are difficulties in the process of technology application and extension. The three-linear analysis algorithms have a high computational complexity, and the computational performance of general embedded processor is not high enough. Because of this, it's difficult to use the general embedded processor directly to implement three-linear analysis, which means the application of three-linear analysis is limited by the insufficient ability of general embedded system. Aiming at the application research of the 973 pre-research project" the analysis methodology and innovative application of Three-dimensional Fluorescence Spectrometry", this paper study the embedded software structure and core algorithm module dedicated embedded computing technology in the three-way data array analysis of complex samples.1. Through the research of experimental procedure and data flow of three-linear analysis application, this paper design a dedicated embedded computing structure for three-way data array analysis of complex samples. Through the study on the cross advantages between the structural and data flow characteristics of common three-linear analysis algorithms and the reconfigurable computing based on FPGA, propose a scheme that using the FPGA application technology with reconstruction ability to optimize the computing of three-linear analysis algorithms. This will provide a basis for the development of dedicated computing device and portable equipment for applications of three-linear analysis methodology.2. Taking the widely applicable and the typical tri-linear decomposition algorithm PARAFAC algorithm for example, the paper studies in some key issues for the FPGA design of the algorithm, using the research and development process of AccelDSP. It includes the type selection design of matrix inversion component, the parallel description of algorithm, the model transform from floating point to fixed point, the selection of storage types of variables, etc. The solution and implementation techniques of these issues provide support for the embedded implementation scheme of three-linear analysis based on FPGA. 3. Through implementation and situation of the use of the top level software and hardware system, the paper validates the correctness of the dedicated embedded computing structure; based on the validate flow of Xilinx AccelDSP, the paper completes the simulation verification of the FPGA implementation of PARAFAC algorithm modules, and analyses the simulation results, and compares difference between the embedded implementation based on FPGA and the software implementation on general computing platform. The results show that using the FPGA technology to implement the three-linear analysis algorithm can make full use of the inherent parallelism of algorithm, and make the computing more efficient.This dissertation research contributes significantly to furthering the application of the embedded system in analytical science, accelerating the spread and use of the three-linear analysis methodology, improving the intelligence of analytical instrument and the automation of experimental procedure.