A pilot scale reverse flow reactor for catalytic combustion of volatile organic compounds (VOCS) in contaminated air is studied and modeled. The quasi-steady state model of temperature profile for the reverse flow reactor is developed in terms of RBF (Radial Basis Function) neural networks. The deep knowledge repository with respect to temperature profile is yielded based on the determinant mathematical model, which increases the 'extrapolability' and 'reliability'. Additionally, the model's accuracy is improved by adjusting the model parameters advisably. For predicting and controlling the transient temperature, a real-time prognostic model of temperature profile is built based on dynamic RBF neural networks by using the Time Delay Neural Network (TDNN), which is to save the previous state in the time-delay cell. Simulation results have proved that the models presented in this paper are simple, highly accurate and can satisfy the control requirements.