Accurate tool wear modelling is indispensable for successful hard turning technology implementation. In this study, a Hybrid Neural Network-based modelling approach, which integrates an analytical tool wear model and an artificial neural network, is proposed to predict Cubic Boron Nitride (CBN) tool flank wear in turning hardened 52100 bearing steel. Extended Kalman Filter algorithm is used to train the proposed neural network, and the network connectivity is further optimised to achieve an improved and robust modelling performance. Results show that the proposed Hybrid Neural Network excels the analytical tool wear model approach and the general neural network-based modelling approach.