Monitoring structural impact is one of the main research aspects of structural health monitoring. A new algorithm for impact localization(IL) is proposed in this paper through a two-dimensional beam-focusing (TDBF) probabilistic tomography theory. The newly developed method is based on the arrangement of cross-shaped piezoelectric sensor (PZT) which has greatly improved the accuracy of IL. Moreover, in the proposed TDBF localization algorithm, the accuracy of the ultrasonic wave velocity achieved during testing has emerged as the main factor that ameliorates the IL. The probability distribution of the velocity is obtained through statistical analysis and introduced into the TDBF localization algorithm to localize the impact through probabilistic tomography. To further improve the accuracy of the IL, Bayes theory is also adopted to fuse two impact tomography obtained by horizontal sensor array and longitudinal sensor array, respectively. The effectiveness and credibility of the TDBF-IL are verified through comprehensive experimental research which indicate the promising prospects of proposed method in accurate and effective monitoring of structural health.