The American Society of Civil Engineers (ASCE) have built a scale model of a four story building and made available the acceleration data for the undamaged structure and for eight different configurations of damage which is used to assess techniques for damage detection, damage severity assessment, damage localization and identification of damage type. Data representing Ambient, Shaker and Hammer as input excitations are available. Previous publications that have used the experimental data have reported limited success in the detection and damage severity assessment. In this paper we describe a methodology that uses a combination of wavelet multi-resolution analysis, principal component analysis, the Welch spectrum estimation method and Adaptive filtering to track changes in natural frequencies. We have successfully detected, for all three input types, all of the damage configurations and completed a severity assessment. This is achieved without explicitly using the input measurements, a mathematical model of the structure or measurements from the damage cases.