Random vibration test was done on aluminum honeycomb sandwich board. The result suggested that chaotic behavior is found in the test data. The Volterra expression can not be gained from the input and output data directly in this paper. The state space reconstruction was used to convert the system observed data into the quasi-input/output pairs, and the second-order Voherra adaptive filter was used to predict the test data. It is shown that combining the state space reconstruction with the Volterra adaptive filter, these chaotic series could be accurately predicted.
