An intelligent capacitance level measuring technique using optimal ANN

This paper aims at designing an intelligent level measurement technique by Capacitance Level Sensor (CLS) using an optimal Artificial Neural Network (ANN). The objectives of the present work are to (i) extend the linearity range of measurement to 100% of the full scale, (ii) make the measurement technique adaptive of variation in (a) permittivity of liquid, (b) liquid temperature and, (iii) to achieve (i) and (ii) using an optimized neural network. An optimized ANN is considered by comparing various schemes, algorithms, and number of hidden layers based on minimum mean square error (MSE) and Regression close to 1. The output of CLS is capacitance. A data conversion unit is used to convert it to voltage. A suitable optimized ANN is added, in place of conventional calibration circuit, in cascade to data conversion unit. The proposed technique provides linear relationship of the overall system over the full input range and makes it adaptive of variation in liquid permittivity and/or temperature. Since, the proposed intelligent level measurement technique produces output adaptive of variations in liquid permittivity and temperature, it avoids the requirement of repeated calibration every time the liquid under measure is replaced or there is any variation in liquid temperature. ANN is trained, tested and validated with simulated data considering variations in liquid permittivity and temperatures. All these variations are considered within specified ranges. When an unknown level is tested with an arbitrary liquid permittivity and temperature, the proposed technique has measured the level correctly. Results show that the proposed scheme has fulfilled the objectives.