The measurement of vibrations of the turbomachinery is usually executed by a strain gauge. Since the turbine has multiple blade stages, it is generally difficult to install the strain gauges on the blades and retrieve the signals especially in the intermediate stages. Also, it is important to secure installation of lead or copper wiring against severe conditions such as steam moisture drain and elevated temperature resistivity. A non-intrusive vibration measurement technique called as Blade Tip Timing (BTT) has been developed to cope with these problems. Some laser beam sensors (called probes) are installed in the casing of the turbine assembly. These sensors measure the time of each blade tip passing under the reflection of the sensor laser beam. This time signal, referred as time of arrival (TOA), serve as the input data for the BTT. Several BTT analysis methods have been developed so far. But only a few have satisfactory results. Auto-Regressive analysis methods developed by Dimitriadis and Galligo Garrido have shown better results so far. In AR method, every time signal is a sum of all the signals measured by the preceding probes.
We solve the AR system to get frequencies and amplitudes of the vibration of the blades. But when it comes to noise in the real system these methods do not behave very accurately. This work is an attempt to increase the accuracy of the Auto-Regressive methods. Our aim is to include random noise and study the effects of noise on the system.