This project investigates online continual learning methods for sequential and non-stationary data. The main objective is to develop and evaluate learning systems that can continuously adapt to new data distributions, operating conditions, tasks, and classes while reducing catastrophic forgetting. The research covers conventional neural networks and spiking neural networks, as well as different continual learning strategies such as experience replay, memory management, regularization, dynamic model adaptation, and online representation learning. Experiments will be conducted using public image, sensor, and industrial time-series datasets, including bearing vibration data. Performance will be evaluated in terms of classification accuracy, forgetting, computational efficiency, memory consumption, robustness, and adaptability. The requested computational resources will support model training, systematic comparisons, hyperparameter evaluation, ablation studies, repeated experiments, and reproducibility checks.