Prediction of Flow through Proportional Control Valves with a Fixed Restriction

Predicting gas flow is difficult because each type of gas has different properties that affect how fast, slow, or even the path the gas medium takes while flowing. In this research, the focus will be predicting gas flow through a proportional control valve with a fixed restriction in a flow control device. Research has shown that artificial intelligence is Excel ® is very capable at using parameters of difficult-to-model relationships, interpreting, and organizing the parameters into a usable model (Haykin, 2009). This study aims to use artificial neural networks (ANN) to model output flow using a proportional control valve with a fixed restriction in a mass flow control device. In achieving this aim, the industries that use these valves will be able to increase the safety, accuracy, and performance of processes that use proportional valves. To accomplish this aim, three experiments were conducted and analyzed, to validate the three goals set forth; determine architecture of an ANN, determine performance of trained ANN, and demonstrate the ability of ANN to handle variations from design set point. In accomplishing these three goals through three experimental setups, it is determined that by utilizing ANN’s, an accurate model of proportional control valves with a fixed restriction can be produced. Further research needs to identify other areas where ANN’s could potentially be applied to increase performance in hard to model processes