Small scale ultrasonic piezoelectric actuators performance strongly depends on not well-known contact dynamics. Deep Neural Network (DNN) sees their use in physic simulation growing as their flexibility allows better performance especially when dynamics laws are yet to be explored. A Deep Learning approach for contact is presented, motivated and tested. The focus of this paper is on normal contact prediction, providing the basis to a complete study including both normal and tangential force. After existing friction models are presented, a real world test bench is introduced along with its digital twins. It provides data for the training and validation of a deep Reinforcement Learning (RL) model.