Proceedings of COBEM 2011 Copyright c  2011 by ABCM 21st International Congress of Mechanical Engineering October 24-28, 2011, Natal, RN, Brazil SENSITIVITY OF VIBRATION SIGNALS TO DETECT EARLY SPALLING STAGES, BY COMPARISON TO ARTIFICIAL DEFECTS INDUCED BY SPHERICAL INDENTATIONS Adriano Gonçalves dos Passos, adriano.utfpr@gmail.com 1 Cesar Ricardo de Oliveira, cesar.ricardo@gmail.com 1 Cristiano Brunetti, cristiano.brunetti@renault.com 2 Carlos Henrique da Silva, carloshs@utfpr.edu.br 1 Giuseppe Pintaúde, pintaude@utfpr.edu.br 1 1 Laboratório de Superfícies e Contato (LASC) – Universidade Tecnológica Federal do Paraná (UTFPR) 2 Renault do Brasil S.A. – Department of Materials (DEVM/DIMat-A) Abstract. The detection of premature failure in components subjected to rolling contact fatigue is very important and the vibration signal analysis is a powerful technique to detect the early stages of spalling. This study aims to verify the sensitivity of this methodology by comparing the results obtained evaluating the vibrational patterns in steel specimens with an artificial defect to the results obtained by measuring the signals from an austempered ductile iron (ADI). For that, it was chosen the testing model ball-on-flat. The shape of artificial defects used as a vibration standard was based on the spalling characteristics observed in ADI at previous studies and literature. Also, the smallest artificial defect produced was related to the average size of graphite nodules, a potential micro-constituent to nucleate cracks. All rolling contact fatigue tests were performed using 3.0 GPa on ADI specimens, the oil ISO 46 at 85 ◦ C was used as lubrificant mean. Piezoelectric accelerometers were used to capture the vibration signals and the signal acquired and processed in a National Instruments LabView platform (NI–LabView). The results showed that, with the current technique, there is a minimum size of artificial defect in which the signal of vibration technique is able to detect. However, this treshold showed to fully compative to the average sizes of spallings. Keywords: Rolling contact fatigue; subsurface cracks; vibration signal analysis; austempered ductile iron. 1. INTRODUCTION Mechanical machinery under long cycles of functioning, unavoidably, suffer, at some point, from fatigue damage (Norton, 2004). However, a severe state of damaging should be avoided to ensure the safety of operators and plant (Vale, 2007). According to Almeida (2005) most of manufacturing plants that are based on the use of mechanical equipments have maintenance scheduled based on vibration signals. Bezerra (2004) states that processing vibration signals via time-domain analysis is relative simple, on the math point of view, reliable, due large amount of samples processed, and fast, usually processed in field by dedicated devices. Root Mean Square (RMS) and Crest Factor (CF) are the most used. Moreover, Martin and Honarvar (1995) demonstrated that more complex statistical analysis provides better detection of damages, for example the fourth standardized moment commonly named Kurtosis. Many researchers have being studying vibration signals with propose of obtaining correlations with wear mechanisms with the vibration behavior of the system. Mori et al. (1996) managed to identify the early stages of pitting in contact fatigue tests utilizing vibration signals and wavelet transform (WTL). Another state-of-art technique that have being used to identify several types of damages in mechanical components is the Acoustic Emission (AE), that technique used with vibration proved to be convergent and reliable to the detection of damage in rolling bearings (Passos et al., 2010; Silva et al., 2009; Choudhury and Tandon, 2000). The study and detection of wear damage on materials with complex microstructure, as ADI or sinterized materials or casted components, using vibration signals is difficult. Due the high level of noise generated from the graphite nodules or even porosity the signal acquired might need be filtered. Also external factors as electric motors and mechanical couplings and components, when operating, generates extra noise to the system, that one in some scenarios being even bigger then the damage signal. Some authors as Chiementin et al. (2007); Mori et al. (1996); Purushotham et al. (2005) utilized advanced techniques to filter noise and isolate the signal, however those techniques usually consume heavy processing capability and time, being less viable in an industrial environment. This paper has as main objective the development of a testing methodology to identify surface damage type spalling. Also, this work oriented to materials with homogeneous structure like steel and non-homogeneous structure like nodular irons.