A critical part of structural health monitoring is accurate detection of damages in the structure. This paper presents the results of two multi-class damage detection and identification approaches based on classification using Support Vector Machine (SVM) and Artificial Neural Networks (ANN). The article under test was a fiber composite panel modeled by a Finite Element Model (FEM). Static strain data were acquired at 6 predefined locations and mixed with Gaussian noise to simulate performance of real strain sensors. Strain data were then normalized by the mean of the strain values. Two experiments were performed for the performance evaluation of damage detection and identification. In both experiments, one healthy structure and two damaged structures with one and two small cracks were simulated with varying material properties and loading conditions (45 cases for each structure). The SVM and ANN models were trained with 70% of these samples and the remaining 30% samples were used for validation. The objective of the first experiment was to detect whether or not the panel was damaged. In this two class problem the average damage detection accuracy for ANN and SVM were 93.2% and 96.66% respectively. The objective of second experiment was to detect the severity of the damage by differentiating between structures with one crack and two cracks. In this three class problem the average prediction accuracy for ANN and SVM were 83.5% and 90.05% respectively. These results suggest that for noisy data, SVM may perform better than ANN for this problem.