Research On The Parallel Algorithms In Acoustic Field Time Domain Signal Prediction

With the development of science and technology, the processing capacity of singleprocessor computer is being increased year by year, but due to the limit of technical level, thesingle processor computer can hardly satisfy the demand of computing resources forchallenging large-scale computing subject in many fields, so in addition to increase thecapacity of the processor, parallel computing is a kind of effective method to improve thecomputational capacity.Parallel computing technology has been applied in many fields. if it is difficult to rely onmodifying model itself to reduce the computing time, we can depend on the parallelalgorithm to solve the problem. This paper applied parallel computing technology to theexisting sound field calculation model, in order to solve the promble of computationalcomplexity, large amount of calculation, slow computation speed in underwater acoustic field,a certain amount of exploration for the realization of the sound field to calculate the real-timerequirements. This paper makes researches as follows:(1) Studied Parallel computing hardware platform, mainly including the existing parallelcomputer system structure, and the language based on CUDA parallel platform GPU; theparallel computing language environment, mainly including the existing parallel language,MPI programming environment, and CUDA language programming model, storage modeland so on.(2) The calculation of the acoustic field model is the basis of Sound field forecast,thispaper studied the sea of the sound field modeling methods, made a deep research on thenormal modes KRAKEN model, calculation principle and the realization of the broadbandserial KRAKEN principle. This paper introduced the establishment of the pulse propagationmodel in time domain of the two methods, and the frequency domain solutions of Fouriersynthesis method.(3) Introduced the using and installation method of message transfer model MPI, CUDAstandard, set up a parallel computation of operation environment; Realized based on theKRAKEN MPI parallel computation, broadband acoustic CUDA CUFFT library based on therapid IFFT parallel algorithm, and finally by combining the two parallel algorithm to realizethe broadband sound field in the received signal time domain parallel forecast simulation.(4) Through different amount of data test show that the parallel program could guaranteethe serial program with the calculation precision of the effectively reduce calculation time, improve operation efficiency.