Study relating Acoustic Emission (AE) energy to fracture energy has been conducted in the past and a correlation has been reported between the two. The scope of these studies has remained limited to laboratory size specimens with known crack location and the energy release being investigated on a global scale. The objective of this research is to track the local energy release due to cracking in concrete using AE monitoring and understand the relationship between AE energy and crack energy estimated using nonlinear FE model. This will allow to keep the track of energy release due to cracking and use AE energy as a measure for the structural health condition. The study involves challenges with respect to AE source identification, estimating AE energy at source location accounting for attenuation losses and estimating local energy in the numerical model. These issues are discussed in detail in this thesis and the use of AE in crack monitoring is critically examined. In the first part of the thesis, AE source classification methods including signal-based approach and parameter-based approach are reviewed. The classification methods help distinguish the AE activities due to crack opening from ones related to friction. An approach for signal-based AE classification using the AE signal in the frequency domain is proposed. This approach is then compared to existing bivariate and multivariate parameter-based classification methods. In addition to this, a novel partial power-based method for AE source classification is also proposed. The existing parameter-based classification methods are found to have a similarity of less than 50% in case of bivariate methods and a little over 50 % in the case of the multivariate method when compared to the signal-based method. This is because these methods are unable to notice small differences in AE signals. On the other hand, the partial power-based method has a similarity of about 75 % to the signal-based method. In addition to this, the partial power-based method is much faster than the signal-based approach, thus providing a good alternative to the existing AE classification methods. In the second part, attenuation in AE signals is studied. Experiments on sound concrete and cracked concrete have been performed to study the attenuation in concrete media and through a crack, respectively. AE attenuation due to elastic wave propagation is made under the assumption of a Rayleigh wave and the material attenuation factor (α) is estimated to be 2.473 m-1. Crack attenuation factor (C.A.F.) is introduced to determine the energy loss through a crack. Auto Sensor Test (AST) measurements made during the experiment were used to estimate C.A.F. AST measurements are found to be sensitive to the strain changes within the concrete and are thus able to predict the occurrence of the crack in advance. In the last part, a methodology to estimate the local energy release in the numerical model is proposed and then verified using a notched beam as a test model. A rotating crack approach for modelling is adopted with tension behaviour defined using the Hordijk curve. The proposed methodology is applied to the girder model to estimate the energy released locally. The numerical energy trend thus calculated is compared to the AE energy trend at the crack location. The AE energy predicts the occurrence of the first flexure crack at 90% of the cracking load as per numerical energy. A possible explanation for this is that AE can also detect the presence of the microcracks, which the current numerical model cannot. On comparing the estimated energies released due to AE and numerical model in the flexure zone it can be concluded that the relationship between the AE energy and numerical energy is non-linear. Local energy release trend for AE and the numerical model with increasing load is similar when the flexure cracks are generated, although slight deviations start to occur when the shear crack is created.