Transient dynamical analysis of strain signals in sheet metal stamping processes

On-line monitoring of stamping processes can be carried out based on various sensors, such as force, strain, acceleration, proximity, and acoustic emission sensors. The strain sensor signal is the most favourite because of its effectiveness and acquisition cost as well as it contains rich information about the stamping process. The key problem of stamping monitoring is how to extract features from the strain signal to effectively detect the faults. The strain signal, however, is a transient signal that depends on many factors. In this paper, it is intended to address some new methods to analyse the transient strain signal with the objective of decomposing it in order to understand the dynamics of the stamping process and extract a malfunction signal for fault detection. A latent process model method, which is a combination of a time-varying auto-regression model and a dynamic linear model, is initially presented. Continuous wavelet transforms and a new discrete wavelet transform (maximum overlap discrete wavelet transform) are then addressed to project the transient signal into a time scale plan to represent the dynamical behaviour in a different way. Empirical mode decomposition is finally employed to decompose the transient signal into a finite and often small number of intrinsic mode functions (IMF). The advantage of this new method is that it is adaptive and highly efficient. The performance of the methods employed in this paper is reviewed using two real strain signals in a sheet metal stamping process. It is found that these methods can efficiently provide the energy–frequency–time distribution of the transient strain signal.