Thermoacoustic instabilities in combustion chambers represent a serious threat to combustion systems, which can lead to performance degradation as well as to relevant structural damages. The nonlinearity of these phenomena represents a serious obstacle to the prediction of the evolution of the relevant system variables. The early prediction of out-of-control states in combustion chambers might represent an important step ahead in the design of accurate control system for the suppression of undesired behaviours. This study proposes the application of control charts to the prediction of out-of-control states in an experimental combustion chamber. EWMA control charts have been used because they are very useful when on line single measurements are collected from the process. In order to deal with the high level of autocorrelation characterising the deterministic nonlinear experimental measurement, the EWMA control charts have been applied to the residuals of an input-output NARMAX identification model, implemented by means of a Multilayer Perceptron artificial neural network. Obtained results show the ability of the control charts in detecting unstable combustion phenomena, pointing out the promising application of these statistical tools in the diagnostic of combustion instabilities. (C) 2008 Elsevier Ltd. All rights reserved.