Neural modeling for cytochrome b5 extraction

In this work a neural model for cytochrome b5 extraction in batch and in continuous operation was developed. The best feedforward arquiteture achieved for batch operation modeling was 3-4-1 and 3-8-2 for the continuous operation. It was observed that among the models developed, the best adjustment was that obtained with Bayesian regularization algorithm training. Deviations of less than ±10% were observed for the experimental data and they are similar for the neural model, since it was statistically proved the null hypothesis in the comparison between the two independent samples (experimental and predicted outputs).