This research introduces an innovative Deep Neural Network (DNN)-based framework for managing autism spectrum disorder (ASD) incidents, such as seizures and sensory meltdowns in children. The proposed system incorporates wearable devices, hybrid CNN-LSTM architectures, and cloud storage to enable real-time prediction, self-regulation, and physician collaboration. A comprehensive Failure Modes and Effects Analysis (FMEA) is conducted to improve system dependability and reduce hazards. Assessment utilizing the CHB-MIT EEG Seizure Dataset and AUT-PRE-Behavioral Dataset reveals an accuracy of 96%, with additional recommendations for further improvements.