This paper presents a novel control method for unbalanced normal force (UNF) reduction in the eccentric double-sided linear switched reluctance machine (DLSRM). Online detection of the DLSRM eccentricity is investigated via magnetic circuit analysis as well as finite element method. It is shown that the DLSRM eccentricity is directly proportional to the difference between reciprocal values of winding unsaturated inductances on both sides of the DLSRM mover. In order to improve the accuracy of the eccentricity detection, a new scheme to estimate the winding inductances is proposed on the basis of the voltage pulse injection considering iron losses in the test DLSRM in the laboratory. Afterward, a control algorithm for the UNF reduction, according to the estimated real-time eccentricity, is designed, and the corresponding power converter and flow chart are illustrated in detail. To verify the proposed methods, a prototype experimental platform is developed, and the experimental results confirm that the proposed methods in this study are feasible and effective.