Epilepsy is one of the most prevalent neurological disorders among children. The study of surface EEG signals in patients with epilepsy by techniques based on symbolic dynamics can provide new insights into the epileptogenic process and may have considerable utility in the diagnosis and treatment of epilepsy. The goal of this work was to find patterns from a methodology based on symbolic dynamics to characterize seizures on surface EEG in pediatric patients with intractable epilepsy. A total of 76 seizures were analyzed by their pre-ictal, ictal and post-ictal phases. An analytic signal envelope algorithm was applied to each EEG segment and its performance was evaluated. Several variables were defined from the distribution of words constructed on the EEG transformed into symbols. The results showed strong evidences of detectable non-linear changes in the EEG dynamics from pre-ictal to ictal phase and from ictal to post-ictal phase, with an accuracy higher than 70%.