An Intelligent Flow Measurement Technique Using Neural Network

Abstract

A variable area type flow meter called rotameter, a measuring instrument in which position of the flow is indicated by the stainless steel float which is present in rotameter. In the process industries the readings will be required in control room hence the float is converted into voltage. For non-contact flow measurement a hall effect sensor is kept on the rotameter which acts as secondary sensor. To convert float into voltage the hall effect sensor sense the magnetic field created by the magnet kept on the float, as water flow through the rotameter the float increases and the magnet varies its position and the hall sensor sense the change in magnetic field created by the magnet. The hall voltage varies with different parameters which affect the hall voltage measurement. As temperature is very slow process and water flow is very fast process both are considered for our work as temperature varies non linearly with voltage. The hall sensor output is amplified by instrumentation amplifier and that is given as input to the ANN. For neural network 70% readings has to be used for training and 30% readings should be used for testing process. Error is given back to the input so back propagation algorithm is used. By weight adjustment ANN gives the required output with iterations so we will get the accurate output.