Abstract: In view of the concentration of measurement points in the traditional sensor optimization method with a large number of configurations， the relationship between optimization results and prediction errors was analyzed based on the information entropy theory， and an improved optimization method considering prediction errors was proposed. The reciprocal of the sum of normalized modal strain energy was used as the principal diagonal element of the prediction error covariance matrix， and the non-diagonal element was expressed by an exponential correlation equation combining the distance of measurement points and the response level. Aiming at maximizing the determinant of Fisher information matrix， the forward sequential sensor placement method was used to obtain the optimal location of measurement points. The finite element analysis results of the urban rail vehicle frames showed that the improved sensor optimization method is superior to the classical method in the performance of three evaluation criteria， and significantly reduces the concentrating effect of measurement points， which proves the effectiveness of the improved method in the configuration of bogie frame strain sensors.

Key words: urban rail vehicle, bogie frame, optimal sensor configuration, strain mode, information entropy