Structural Health Monitoring (SHM) and Non Destructive Evaluation (NDE) is considered a necessity for optimized lightweight structures. Thus several SHM strategies have been suggested for assessing the condition of such structures. The key aspect of any SHM method is the choice of the damage sensing feature. Ideally the damage sensing feature should be sensitive to small levels of damage, insensitive to measurements noise and ambient condition changes, like loading conditions, temperature, moisture etc. For beam structures the Neutral Axis (NA) is known as a very informative damage sensing feature. The NA position depends on the condition of the structure alone and it is insensitive to the load level. Through the use of advanced data processing tools like the Kalman Filter (KF), the NA estimation can be made robust in the presence of measurement noise and changing ambient conditions. In this paper, the use of a KF based NA estimation technique is investigated for determining the size of a damage. The methodology is employed in experiments on a thin-walled beam with rectangular cross-section under 4-point bending. Cracks of different lengths are introduced in the beam and based on the strain measurements the NA is estimated. The estimated position of the NA is then correlated with the size of the damage.