LINEAR OR NONLINEAR? A BICOHERENCE BASED METRIC OF NONLINEARITY MEASURE

摘要

Most process controllers are linear and do perform well when they are optimally tuned. However, the presence of nonlinearities in a process can affect the control performance adversely, e.g., valve stiction may cause oscillations; other nonlinearities e.g. due to process operating conditions change, may render the current tuning parameters as 'non-optimal' and so on. For this reason a data-based quantification of process nonlinearity is useful. A new Total Nonlinearity Index (TNLI) is defined to quantify nonlinearities. The method is based on a better estimation of the bicoherence with less number of spurious peaks. The previously published Non-Gaussianity Index (NGI) and Nonlinearity Index (NLI) are modified to overcome their limitations. The efficacy of the proposed indices has been shown using two illustrative examples and one simulation example.