Non-linear qualitative signal processing for biological systems: application to the algal growth in bioreactors

We present in this paper a qualitative method to validate and monitor the structure of a non-linear model with respect to experimental data, under some hypotheses. This method is broadly independent of the analytical formulation of the model, and depends only on the qualitative structure (the signs of the Jacobian matrix). The temporal sequences of the extrema of a filtered experimental signal are compared with the transitions allowed by a graph. In particular, we show that the usual moving average of the outputs follows this transition graph. We apply this method to compare models of algal growth in a bioreactor with experimental data.