Switching Learning Law for Differential Neural Observer for Biodegradation Process

In this paper, it is presented a differential neural network supplied with a new learning law based on the sliding mode approach. The state observer is employed to estimate the dynamics states of degradation mathematical model, where the incomplete information and the limited on-line measure problems are considered. A new training method is applied in the learning algorithm is proposed to reconstruct biomass, organic matter recalcitrant concentrations and volume of biological culture evolutions. This allows ensuring an upper bound for the weights time evolution. This new scheme gives the possibility to construct not only one adaptive process but a set of learning laws. The effectiveness of this algorithm is shown by numerical results.