Bifurcation Predication in Axial Compressors with Nonuniform Inflow via Deterministic Learning

Abstract

The nonlinear dynamics of fluid instabilities such as rotating stall and surge in axial compressors are typically modeled as subcritical Hopf bifurcations with hysteresis. The bifurcation prediction provides an effective approach to avoid the occurrence of compressor’s instability. In this paper, based on a fluid dynamic model developed recently, a stall precursor detection approach employing deterministic learning (DL) is proposed for bifurcation predication in axial compressors with nonuniform inflow. The stall precursor near the bifurcation can be obtained as the throttle area parameter approaches its critical (or bifurcation) value. Firstly, the system dynamics underlying normal and stall precursor are locally approximated accurately through DL. The obtained knowledge of dynamics is stored in constant radial basis function (RBF) networks. Secondly, a bank of estimators is built up using the stored constant RBF networks to represent the learning normal and stall precursor patterns. By comparing each estimator with a test system, the average [Formula: see text] norms of the residuals are taken as the measure of the dynamical differences between the test system and the learning patterns. The occurrence of stall precursor as a bifurcation predication signal can be rapidly detected according to the smallest residual principle. Finally, simulation results are given to show the effectiveness of stall precursor detection approach.