In this work it is demonstrated that robust burn control in long-pulseoperations of subignited thermonuclear reactors can be achieved with radialbasis neural networks (RBNNs) composed of Gaussian nodes in the hidden layerand sigmoidal units in the output layer. The results reported here correspondto a volume-averaged zero-dimensional nonlinear model of a subignited fusionreactor with design parameters corresponding to those of the ITER-EDA group.The control actions are implemented through the concurrent modulation of theD-T refuelling rate, a neutral 4He beam and an auxiliary heating power,constrained to lie below maximum allowable levels.It is shown that the resulting network provides feedback stabilization over awide range of energy confinement times for plasma density and temperatureexcursions significantly far from their nominal operating values. The resultsshow that the RBNN feedback-controlled nonlinear system is stable regardlessof any particular scaling law, as long as the confinement time lies within thescope of the training region. In addition, it also shows robustness withrespect to noise in the energy confinement time value fed into the controllerduring simulated transients of a thermonuclear system using a particular ELMyscaling law, as well as with respect to the thermalization time of the alphaparticles produced by fusion.