Abstract
The present paper shows signal processing techniques applied to experimental data obtained from a T100 microturbine connected with different volume sizes.
This experimental activity was conducted by means of the test rig developed at the University of Genoa for hybrid systems emulation. However, these results can be extended to all advanced cycles in which a microturbine is connected with additional external components which lead to an increase of the plant volume size. Since in this case a 100 kW microturbine was used, the volume was located between the heat recovery unit outlet and the combustor inlet like in the typical cases related to small size plants. A modular vessel was used to perform and to compare the tests with different volume sizes.
The main results reported in this paper are related to rotating stall and surge operations. This analysis was carried out to extend the knowledge about these risk conditions: the systems equipped with large volume size connected to the machine present critical issues related to surge and stall prevention, especially during transient operations towards low mass flow rate working conditions.
Investigations conducted on acoustic and vibrational measurements can provide interesting diagnostic and predictive solutions by means of suitable instability quantifiers which are extracted from microphone and accelerometer data signals. Hence different possible tools for rotating stall and incipient surge identification were developed through the use of different signal processing techniques, such as Wavelet analysis and Higher Order Statistics Analysis (HOSA) methods. Indeed, these advanced techniques are necessary to maximize all the information conveyed by acquired signals, particularly in those environments in which measured physical quantities are hidden by strong noise, including both broadband background one (i.e. typical random noise) but also uninteresting components associated to the signal of interest. For instance, in complex coupled physical systems like the one it is meant to be studied, which do not satisfy the hypothesis of linear and Gaussian processes inside them, it is reasonable to exploit these kinds of tools, instead of the classical Fast Fourier Transform (FFT) technique by itself, which is mainly adapt for linear systems periodic analysis.
The proposed techniques led to the definition of a quantitative indicator, the sum of all auto-bispectrum components modulus in the subsynchronous range, which was proven to be reliable in predicting unstable operation. This can be used as an input for diagnostic systems for early surge detection. Furthermore, the presented methods will allow the definition of some new features complementary with the ones obtainable from conventional techniques, in order to improve control systems reliability and to avoid false positives.
Predicting livestock behaviour using accelerometers: A systematic review of processing techniques for ruminant behaviour prediction from raw accelerometer data
