Model Identification and Data Analysis

11Case Studies

11.1 Introduction

The methods of prediction and identification explained in this book have a huge variety of applications.

Here, we focus on two case studies. The first one is the analysis of the data of an earthquake that took place in Kobe (Japan) in 1995. The data cover various phases and can be distinguished in three segments; the first one corresponds to a normal seismic activity, while the third one is the earthquake phase. In between there is a transition phase on which our attention focuses in order to find some feature hidden in data to detect the occurrence of the earthquake. In this case study, input–output models will be used, with focus on parameter estimation and model complexity selection.

The second case study concerns the estimation of the frequency of a sinusoid from noisy measurements. The problem will be tackled both with a prediction error identification method based on input–output models and by extended Kalman filtering state space techniques.

11.2 Kobe Earthquake Data Analysis

Kobe earthquake occurred on 16 January 1995, at 20:46:49 (UTC) and measured 6.8 on the moment magnitude scale. Our analysis is based on the data collected by a seismograph located at the University of Tasmania, Hobart, Australia. Measurements began at 20:56:51 (UTC), with a sampling interval , and lasted for 3000 seconds (about 51 minutes). The corresponding time series ...