Enhancing the linearity of LVDT by two-stage functional link artificial neural network with high accuracy and precision

Abstract

Linear variable differential transformer (LVDT) is an important position sensor for many industrial equipments. LVDT has a nonlinear response in its full range and hence a reduced range is chosen as operability range. However, this linear range can be enhanced by suitably placing a nonlinear function model after the LVDT which has an inverse response. Functional link artificial neural network (FLANN) was recently used to compensate this nonlinearity and was tested with a very low precision. Therefore, in this paper a new process of LVDT nonlinearity compensation is proposed which includes three steps. In first step, a best fit direct model of the LVDT is obtained. Then a lower order FLANN is used to roughly compensate the nonlinearity of the LVDT model. After that another higher order FLANN is used to compensate the left over nonlinearity. This two stage FLANN based inverse model was shown to achieve better measurement accuracy with higher precision.