Expert condition monitoring on hydrostatic self-levitating bearings

• Published in: Expert systems with applications Vol. 40; no. 8; pp. 2975 - 2984
• Main Authors: Garcia, Ramon Ferreiro, Rolle, José Luis Calvo, Gomez, Manuel Romero, Catoira, Alberto DeMiguel
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier Ltd 15.06.2013; Elsevier
• Subjects: Applied sciences; Artificial intelligence; Bearings; Bearings, bushings, rolling bearings; Computer science; control theory; systems; Condition monitoring; Connectionism. Neural networks; Drives; Exact sciences and technology; Fault diagnosis; Feedforward neural networks; Fluid dynamics; Fluid flow; Functional approximation; Fundamental areas of phenomenology (including applications); Industrial metrology. Testing; Mathematical analysis; Mechanical engineering. Machine design; Neural networks; Physics; Solid mechanics; Structural and continuum mechanics; Tasks; Vibration analysis; Vibration, mechanical wave, dynamic stability (aeroelasticity, vibration control...); Virtual data; Condition monitoring; Fault diagnosis; Feedforward neural networks; Virtual data; Functional approximation; Vibration analysis; Measurement; Viscosity; Expert system; Dynamic analysis; Dynamical system; Function approximation; Task analysis; Pumping; Accelerometer; Shaft; Feedforward; Signal conversion; Monitoring; Supervision; Rotating machine; Task scheduling; Knowledge base; Neural network; Vibration amplitude; Pressure; Operating conditions; Data conversion; Hydraulic drive; Fault diagnostic; Functional network; Structural analysis
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2012.12.013

► Neural network based measurements. ► Virtual measurements based faults detection. ► Rule based machine protection. ► Virtual measurements of machine disturbances. ► Disturbances identification using virtual vibration analysis. Neural network based functional approximation techniques associated with rule based techniques are applied on the condition monitoring task of rotating machines equipped with hydrostatic self levitating bearings. Based on fluid online measured characteristic data, including pressures and temperature, the inherent hydraulic pumping system and the self levitating shaft is monitored and diagnosed applying vibration analysis carried out using virtual measurements. Required signals are achieved by conversion of measured data (fluid temperatures and pressures) into virtual data (vibration magnitudes) by means of neural network functional approximation techniques. Previous to the condition monitoring task (vibration analysis), a supervision task of the system behaviour is carried out in order to validate the information being processed. It is concluded that the vibration analysis based on the analysis of the dynamic behaviour of oil pressure (non accelerometer based signals) subjected to disturbances such as changes in oil operating conditions including viscosity, is successfully feasible.