METHODOLOGY FOR AUTOMOTIVE AIR-CONDITIONING CONTROL OPTIMIZATION USING ARTIFICIAL NEURAL NETWORKS

The transient nature of automotive air conditioning systems control is generally achieved through proportional–integral–derivative controllers (PID’s) parameters tunning. Due to the vast database available from decades of automotive manufacturers design and expertise, Artificial Neural Networks (ANN) might be able to identify underlying patterns to predict and properly tune the air-conditioning PID control systems under different thermal requirements. Recently, advances in computational capability have enabled compact embarked systems to rapidly solve complex, multi-variable sets of equations, thus allowing for ANN to promptly calculate tunning parameters and act upon PID controllers. As any new application, technical literature is still scarce. On this research, a coupled PID and 6-layers perceptron ANN system was devised, programmed and used to simulate how an air-conditioning system performance can be optimized through proportional–integral–derivative controllers tuning. This proposed setup response was then compared to a conventional heuristic PID tunning method (Ziegler Nichols) to demonstrate how ANN’s can more rapidly stabilize the system output