Experimental Identification of a Mechanical Structure with Characteristic Matrices : Proposed Method with Input Data in Which Noise is Included : Vibration, Control Engineering, Engineering for Industry

抄録

The authors propose a new method for identifying characteristic matrices of mass, damping and stiffness in this paper. The authors have already proposed a method of identification in their first report, but accurate characteristic matrices cannot be identified by this method when errors are included in the transfer functions used as the input data. Some errors are always included in the actual experimental data. Therefore, a method of identification, using actual experimental data, is proposed, based on statistical considerations. In this new method, it is assumed that the error in the input data has a normal distribution. Two numerical examples are shown to demonstrate that the new method can accurately identify characteristic matrices with input data which includes normal distributive error.