Tesis:
A method for the very early detection of rotor-casing rub in aeroderivative gas turbines
- Autor: SILVA BERNÁRDEZ, Alejandro
- Título: A method for the very early detection of rotor-casing rub in aeroderivative gas turbines
- Fecha: 2020
- Materia: Sin materia definida
- Escuela: E.T.S. DE INGENIEROS INDUSTRIALES
- Departamentos: MATEMATICA APLICADA A LA INGENIERIA INDUSTRIAL
- Acceso electrónico: http://oa.upm.es/66046/
- Director/a 1º: ZARZO ALTAREJOS, Alejandro
- Director/a 2º: MUÑOZ GUIJOSA, Juan Manuel
- Resumen: The flawless operation of rotating machines such as turbomachines is key in many branches of Industry. During the last decades there was an increase in the net efficiency and performance of these machines, attained through a minimization of component mass, size and dimensional tolerances, rotor-to-stator clearances; the use of cutting-edge materials with top-class mechanical and thermal properties and the introduction of the newest cooling and lubrication technologies. However, higher performance came inevitably with a trade-off in reliability, since new designs and technologies made the turbomachines more vulnerable to malfunction. A breakdown during operation can carry over unassumingly high economical or even human losses, which justifies the convenience of preventive maintenance programs which allow for early malfunction detection before it could cause over time forced shutdown or even catastrophic failure. One of the most common failures in rotating machinery is rotor-casing rub, which is caused by a combination of excessive vibration with abnormal rotor-stator clearances due to thermal expansion and assembly or maintenance errors. This phenomenon causes wear and overheating and, when unnoticed, may be highly destructive. In most gas and steam turbines, rub is easily detected with displacement sensors mounted between rotor and stator. However, in machines such as aeroderivative gas turbines, acceleration sensors on their casing are the often only means available for condition monitoring due to lack of space inside the machine and the very high stiffness of the bearings. Few published works deal with the detection of rub in turbomachines with acceleration sensors, and nothing has been found on early fault detection of rub with accelerometers. Since aeroderivative turbines are increasingly relevant to sectors such as energy generation and chemical processing this problem offers a fertile ground for active research with important industrial outcomes. The aim of this Doctoral Thesis is the development of a signal-processing methodology for the very early detection of rotor-casing single-point rub in aeroderivative gas turbines with accelerometers on the casing, most often the only means available for their monitoring. This methodology has been devised with the assistance of a mathematical model of an aeroderivative gas turbine. When integrating the model equations—derived with the Finite-Element method—the accelerations of selected casing nodes under either rub of no-rub are available for post-processing and a range of detection methods can be tested and compared. The most optimal procedure must detect very weak rotor-stator contact at very small time scales and be implementable in real time. This novel methodology was then validated in an experimental rotor rig with flexible casing monitored with both accelerometers and proximity sensors. The research found that the Wavelet Synchrosqueezed Transform—WSST, a ContinuousWavelet Transform with energy reassignment—complied with all the requirements. Its efficiency is even more outstanding when acceleration signals are contaminated with noise.