传感器非线性输入输出信号反演处理中的'归十'算法

摘要

The reversal development of input verse output non-linear signals is presented by using a so called "Ten-Normalized Algorithm". The key point is that the maximum and minimum output signals are taken respectively as ten and zero on the ordi-nate axes.Other output signals are located in the range of them according to their ratios. The input signals can be expressed as a polynomial of the reversal development for output signals with five orders, which is an interlaced series with five orders. The approximate coefficients of the first and second order can be obtained from two sampling values by solving their two coupled equations. The other three coefficients of the high orders can be obtained by introducing them into the polynomial of reversal development until each computed values ate respectively close to their sampling values after a lot of iterations from their possible minimum to maximum coefficients of the high orders, step by step. The possible range is estimated by using a normalized matching polynomial,according to all of situations for the probable experimental data distribution to minimize the range for searching the optimum coefficients. At last, a polynomial with five orders of the reversal development for output signals verse input signals is obtained, after the "ten-denormalized"process. An example was taken from the reversal development of input verse output signals for an ethanol sensing device.The input signals were ethanol equi-quantities; the output signals were the resistance of the sensing element or the voltages over it. After processing the output signals through a microprocessor, the ethanol quantities were revealed as results from the reversal development. The precision of the reversal development was ±1.5% of ethanol quantities.%介绍一种传感器中非线性输入输出信号反演的所谓"归十"算法.其要点是将最大和最小输出信号分别作为坐标轴的十和零.其它输出信号按比例位于其中.输入信号可表示成"归十"的输出信号的五阶反演多项式,即五阶交错级数.该多项式的一阶和二阶项系数的近似值可从两组样本值解其联立方程而得到.其它高阶项系数可从它们可能的最小值到最大值逐步代人多项式,多次迭代后使各计算值接近相应样本值而得到.这些高阶项系数最小值至最大值的范围可从大多数试验的各种可能数据分布情况利用规范化多项式拟合法估计得到,以使寻找最佳系数的范围缩小.最后又进行输出信号的"去十"化处理,得到输入与此输出信号的五阶反演多项式.以酒精敏感器件的输入输出信号反演为例加以说明.输入信号是等量的酒精,输出信号是敏感元件的电阻或其上的电压降.输出信号经微处理器处理后,酒精量便以反演结果而被显示.反演的精度为酒精量的±1.5%.