The presence of direct current (dc) biased magnetic feld enhances the vibration and noise in the transformer cores due to magnetostrictive efect in the grain-oriented electrical steel laminations. Based on an Epstein frame, an experimental setup was built to measure the in-plane magnetostriction, the principal magnetostriction was computed under a dc bias, and the infuence of dc bias magnetic feld on the magnetostrictive property of electrical steel sheet was investigated. A model based on a back propagation neural network (BPNN) method assisted with the Levenberg–Marquardt algorithm was proposed to describe the magnetostrictive behaviour in the presence of dc bias. Finally, the BPNN model was verifed by comparing the measured magnetostriction and the computed one. This research is helpful to make an accurate estimation to deformation of iron core under a dc bias.