Non-linear signal analysis applied to surface wear condition monitoring in reciprocating sliding testing machines

When the surfaces of two elastic bodies present relative motions under certain amount of contact pressure the mechanical system can be unstable. Experiments conducted on elastic bodies in contact shown that the dynamic system is self-excited by the non-linear behavior of the friction forces. The main objective of this paper is to estimate the friction force using the vibrations signals, measured on a reciprocating wear testing machine, by the proposed non-linear signal analysis formulation. In the proposed formulation the system global output is the sum of two outputs produced by a linear path associated in parallel with a non-linear path. This last path is a non-linear model that represents the friction force. Since the linear path can be identified by traditional signal analysis, the non-linear function can be evaluated by the global input/output relationships. Validation tests are conducted in a tribological system composed by a sphere in contact with and a prismatic body, which has an imposed harmonic motion. The global output force is simultaneously measured by a piezoelectric and by a piezoresistive load cells. The sphere and prismatic body vibrations are measured by a laser Doppler vibrometer and by an accelerometer respectively. All signals are digitalized at the same time base and the data is transferred to a microcomputer. The non-linear signal analysis technique uses this data to identify the friction force. Nowadays the analysis of nonlinear dynamic systems and the analysis of nonlinear forces applied to linear systems are increasing of importance. These dynamic systems with nonlinear properties or time dependent properties cannot be analyzed using the theories developed to linear systems. Analysis of these kinds of systems demand complex methodologies to predict its response or even to estimate its dynamical characteristics and models parameters. Even the study of a linear systems could require a nonlinear signal analysis methodology, as shown by the following cases: Measuring nonlinear physical quantities using linear instrumentation; predicting the response of a linear system excited by a nonlinear force; and the behavior of a vibratory system subjected to friction force. Classical signal analysis, used in the study of dynamical system, was developed to stationary random data. However, much of the random data of interest in engineering applications is nonstationary when viewed as a whole. Typical examples of nonstationary data are vibrations signals measured in dynamical systems subjected to nonlinear forces, in dynamical systems with temperature dependent physical parameters, signals obtained from dynamical