Today, civil aviation is facing new challenges in nonlinear flight control design. Recent nonlinear control techniques offer solutions to these challenges but also bring the need for onboard models of numerical aerodynamics coefficients. The requirements on these potentially onboard models are very strong, since they must be accurate, reliable and compact to cope with aeronautical design's golden rules.It appears that neural networks can meet the aeronautical requirements. However, the usual neural networks design tools are neither autonomous nor fast enough for standard industrial use. We developed integrated neural network identification software to create new automated tools needed for aeronautical industrial applications, such as architecture optimisation and maximum statistical error quantification.