Non destructive testing methods are often used in order to detect and classify structural flaws. The detection of structural flaws is useful for maintenance. In this paper we propose to classify flaws in ferromagnetic materials by measuring Eddy currents. Our approach consists of two steps. First, we use a system identification algorithm to find a dynamical system which describes the data. Then, we use the parameters of this dynamical system as a feature vector and we use support vector machines in order to classify the various cracks. We test our method on a well-known benchmark.