Robust complex non-linear regression method for the estimation of equivalent circuit parameters of the thickness-shear-mode acoustic wave sensor

Fluctuation in characteristic parameters of the equivalent circuit for the thickness-stear-mode (TSM) acoustic wave sensor is a troublesome problem en countered in their practical applications. It is due to interference from n ormal and non-normal noise of an impedance analyzer. A robust complex non-l inear regression, called complex least trimmed squares regression (CLTSR), is described and utilized in the parameter estimation to alleviate the fluc tuation in this article. The results for simulated data indicated that, whe n the noise distribution was not normal, the CLTSR yields the more robust r esults than the ordinary complex least squares regression (OCLSR), when the noise distribution was normal, almost the same estimates can be achieved w ith these two regression methods. The results for the real data indicated t hat, the noise in experiment could not be regarded as normal distribution. The CLTSR can eliminate the system error and alleviate the parameter fluctu ation, especially fluctuation in L-m and C-m. The robust method, therefore, provides a safe alternative to OCLSR and other widely used methods. (C) 19 99 Elsevier Science B.V. All rights reserved.