Development of a virtual curve tracer for estimation of transducer characteristics under the influence of a disturbing variable

This paper presents an artificial neural network (ANN)-based novel virtual curve tracer (VCT) for estimation of transducer response characteristics under the influence of a disturbing variable for computer-based measurement systems. The disturbing variable effect on transducer output response is a typical problem that affects the accuracy of such systems. Especially, change in transducer excitation causes its response characteristics to be highly nonlinear and complex signal processing is required to obtain its accurate direct model. The proposed VCT used a multilayer feed-forward back-propagation artificial neural network (MLFFBP-ANN)-based two-dimensional (2D) model for accurate fitting of transducer characteristics to measured data under the influence of a disturbing variable. The proposed model is trained with Levenberg-Marquardt learning algorithm for achieving an extremely fast convergence speed as compared to the existing ANN-based techniques.