The present work explores the potential of pressure and radiation sensors for the advanced monitoring/control of industrial flames. These instruments are rugged, non-intrusive and non-expensive and might be used in routine plant operation to obtain direct information from the flame. However, further research is needed to assess the existence of relationships among their outputs and operating conditions as well as to define suitable methods for signal processing. Those aspects have been addressed by means of a thorough experimental programme in a model industrial burner. Parametric analysis of flame signals recorded for a broad range of operating conditions revealed that they varied widely with the actual combustion state. In order to perform a systematic study, different correlation techniques were tried. Multiple regression methods provided some insight into mutual influences among different variables, although only in case of linear dependences. Artificial neural networks have been used as a more versatile type of algorithms, suitable for complex functional forms between input and output variables. Remarkably good results were obtained when NO, emissions or some burner settings were estimated from selected features of the flame signals, supporting their applicability for the development of advanced diagnostic methods in combustion processes. (C) 2007 Elsevier Ltd. All rights reserved.