The aim of this research is to propose a new efficient and reliable approach on the field of Non Destructive Testing (NDT), for the characterization of cracks in non-ferromagnetic material by Electromagnetic Acoustic Transducer (EMAT). EMAT is an ultrasonic technique that generates and detects ultrasonic waves in the conductive material without physical contact. The research goes through two principal phases. The first, which is a forward model, is based on Finite Element Method (FEM). The FEM is applied to simulate the EMAT response (output voltage) to the material under test in order to build a database for the inversion tool. The second is the inverse model and depends on the Partial Least Square Regression (PLSR) method, as it is a fast, simple, and accurate inversion tool, in order to estimate the depth and width of the cracks on the surface of non-ferromagnetic materials. PLSR is a dimensionality reduction method which aims to model the relationship between the matrix of independent variables (predictors) (X) and the matrix of dependent variables (response) (Y). The purpose of PLSR is to find the Latent Variables (LV) that have a higher ability of prediction by projecting original predictors into a new space of reduced dimensions.