L4T recently developed a system based on visible, infrared and ultraviolet scattered colorimetry and absorption spectroscopy, which leads to accurate prediction of failure events in gas turbines due to oil degradation 1(particles, corrosion, chemical aging). The advantages of our system arise first of all from the relative hardware low cost and mostly from the application directly on the machineries (online real time monitoring), leading to an effective and dramatic costs reduction from the timing optimization of obligatory oil drains for standard chemio-physical analysis. Another interesting application could be, for example, in quality food certification.
Principle of operation
Trusted analysis is performed on spectroscopic data using the Principal Component Analysis (PCA)2 statistical method, used as source for supervised learning of an Artificial Intelligence (AI) neural network system. After training, whose duration and accuracy is determined by the completeness of the trusted data sample, the system is then ready to work on the field on blind data. Due the purely statistical nature of the PCA analysis, this approach has the great strength to be applicable to any kind of spectrogram. A very interesting source of information for the status of functioning of a turbine or a motor (or probably any kind of mechanical engine), can be given for example from the acoustic noise spectrum produced from the machine in operation. Anyone can understand the
malfunctioning of its own car engine for example by hearing some “strange noise”: an AI supervised trained automated system obviously gives greater sensitivity and higher accuracy, and most of all a reproducible and systematic diagnosis about the machine condition.
1A. G. Mignani, L. Ciaccheri, N. Díaz-Herrera, A. A. Mencaglia, H. Ottevaere, H. Thienpont, S. Francalanci, A. Paccagnini, F. Pavone, Optical fiber spectroscopy for measuring quality indicators of lubricant oils,19th International Conference on Optical Fibre Sensors. Edited by Sampson, David; Collins, Stephen; Oh, Kyunghwan; Yamauchi, Ryozo. Proceedings of the SPIE, Volume 7004, pp. 70045R-70045R-4 (2008).
2A. G. Mignani, L. Ciaccheri, A. A. Mencaglia, N. Díaz-Herrera, P.B. Garcia-Allende, H. Ottevaere, H. Thienpont, C. Attilio, A. Cimato, S. Francalanci, A. Paccagnini, F. S. Pavone, Optical spectral signatures of liquids by means of fiber optic technology for product and quality parameter identification,
Journal European Optical Society - Rapid Publications vol 4 09005 (2009).