Process monitoring is one of the most important problems in modern industrial processes. How to extract the operating features from the measurement data

determine the change of operating region according to the extracted features and infer its effect in real time is the key problem of process monitoring.Extraction of the qualitative features from process measurement signals with wavelet transform has been studied and a model based approach for casual reasoning is proposed to deduce the causality of the extracted qualitative features. The deductive results are applied to real process monitoring. The theoretical analysis and application results show that the proposed method can effectively describe the process behavior knowledge in multi scale and eliminate the virtual character of expert system in describing the time states. Therefore the proposed method improves the reliability of real time process monitoring system.