It is well known that oscillations are a major cause for inferior product quality and productivity losses. Understanding the nature and the phenomena that underlie the oscillations is the first step in mitigating their effect on plant performance. Industrial reality is that multiple oscillations are generally present in the data due to several underlying sources. Detection of oscillations and identification of their time periods are difficult due to the presence of noise in data that might lead to spurious peaks in the power spectrum of the process output. This problem of oscillation detection has received much attention in the literature in recent years. In this paper, an oscillation detection approach that is based on processing of the intrinsic modes that are identified by the sieving process of Empirical Mode Decomposition (EMD) is proposed. The advantages of the proposed method are: (i) ability to detect the presence of single/multiple oscillations and identify their time periods, (ii) ability to provide the amplitude of oscillations, (iii) robustness to noise, (iv) capability to handle nonstationary trends and, (v) ability to provide information about dominant and weak oscillatory modes in the process data. Simulation studies demonstrate the robustness of the proposed approach to noise and its ability to characterize multiple oscillations in the process output. Results obtained from this approach on various industrial case studies are promising and seem to indicate that the proposed technique can be readily implemented in industrial environment.