Neural network controller implementation on a supersonic separator

Supersonic Separator was developed in the last decade to replace the conventional method of removing water content from raw natural gas by chemicals. Its first commercialization was in 2004 on Shell B-11 Platform and still in use up until today. The controller in use currently implemented a PID algorithm to control the position of the shockwave within the separator. This shockwave is essential for the separation process. However, a PID control paradigm is quite inefficient due to the non-linear properties of pressure distribution along the shockwave and the behavior of a shockwave that fluctuates generally around 500 Hz. Implementation of a Neural Network based controller on the system may yield better results in terms of controllability and stability as shown by some research due to its predictive and adaptive characteristic.