This study is concerned with void detection of steel products using electromagnetic acoustic transducer (EMAT). The dynamical behaviors of the pulser-receiver EMAT measurement system are described by a transient eddy current analysis. The ultrasonic wave propagation inside steel products is then given by an initial-boundary elastic problem defined on the steel domain with unknown void parameters. Thus the parameter-to-output mapping is developed by time dependent finite element model with unknown parameters. An inversion methodology is proposed using computational intelligent approach based on the reduced order expansion. Results of computational experiments to demonstrate the efficacy of the proposed method are reported.