2009 年 75 巻 753 号 p. 1295-1303
Modal property estimation of railway vehicles by using a linear prediction model is studied in this paper. The relation between input (excitation force or axlebox accelerations) and output (carbody accelerations) of an actual railway vehicle obtained by stationary or running test is expressed by means of an ARX (Auto-Regressive eXogenious) model. The procedure to extract modal properties is described in detail. Determination of appropriate model order (order of prediction coefficients of the ARX model) is especially discussed from the view of practical use. To employ average prediction errors for two different parts of analyzed data is proposed and their effectiveness to determine the model order is evaluated. The availability for the MIMO (multiple-input multiple-output) problem is also described. It is shown that detailed modal characteristics can be identified successfully by using proposed method from measured data for both of stationary and running test.