CAJ | 학술논문

两相流体具有复杂性的流动特性,图像重建的精度是两相流参数准确测量的基础.针对电阻层析成像系统存在的软场特性、强非线性和不适定性,使得重建的图像质量差、计算时间长等问题,基于代数运算的神经网络,给出了一种基于代数神经网络电阻层析成像图像重建算法.该算法通过建立代数神经网络,以测量的边界电压值作为神经网络的输入,将图像重建转变为一个严格对角占优的线性方程组的求解问题,以达到图像快速、准确的重建目的.通过实验仿真分析,该方法具有收敛速度快、代价低和误差小等特点.
량상류체구유복잡성적류동특성,도상중건적정도시량상류삼수준학측량적기출.침대전조층석성상계통존재적연장특성、강비선성화불괄정성,사득중건적도상질량차、계산시간장등문제,기우대수운산적신경망락,급출료일충기우대수신경망락전조층석성상도상중건산법.해산법통과건립대수신경망락,이측량적변계전압치작위신경망락적수입,장도상중건전변위일개엄격대각점우적선성방정조적구해문제,이체도도상쾌속、준학적중건목적.통과실험방진분석,해방법구유수렴속도쾌、대개저화오차소등특점.
Two-phase fluid has complex flow characteristic and the accurate identification of flow regime is the basis of the accurate measurement of two-phase flow's parameter.There are still many defects such as low reeonstraction quality and low reconstruction speed in image reconstruction algorithm because of soft field characteristic,strong nonlinear and ill-posedness of electrical resistance tomography.This paper puts forward a new image reconstruction algorithm for ERT based on algebraic neural network.This algorithm transforms image reconstruction into a problem of solving strictly diagonal-dominant linear equations.Through the simulation experiment analysis,this method has characteristics such as fast convergence,low cost and small error.