grupo de supervisión y diagnóstico de procesos industriales
Novelty detection of dynamic behavior
One interesting posibility in the field of novelty detection is the detection of changes in the dynamic behavior. The GSDPI has been working on an extension of the idea of analytic redundancy using the SOM of dynamic local models, proposing a visualization procedure that uses a local model SOM previously trained to identify process dynamics under normal (no fault) operation. This can be achieved in three stages:
- Estimate a dynamic model of the current process dynamical behavior, using recent process data -a basic parametric model ARX(p,q) can be used, for instance.
- Retrieve the closest local dynamic model of the SOM (i.e. the best matching model) using a proper disimilarity measure to compare the current model and the SOM models.
- Compute the differences between both models. For instance, the difference between logarighmic frequency responses (Bode magnitudes) could be done, resulting in a residual frequency response. These residual frequency responses could be further represented along time, resulting in a spectrogram-like visualization showing a time-frequency evolution of the dynamic differences.
The application of this idea goes beyond the simple detection of changes in the values of process variables, allowing to monitor changes in the dynamic behavior of the process. This is useful, for instance, in the detection of emerging resonances, gain variations at certain frequencies, or other variations in the frequency response of the process. Our experiments using this method have shown promising results with real data of vertical forces in the rolls of a cold rolling mill, where the chatter effect (mechanical self-excitated vibration of the mill around 120Hz) is visually isolated from the rest of the normal vibrations of the mill (see reference paper and presentation below for details).
- I. Díaz, A. Cuadrado, A. Diez, M. Domínguez, J. Fuertes, and Miguel A. Prada "Visualization of Changes in Process Dynamics Using Self-Organizing Maps". In ICANN 2010, vol.6353 of ''Lecture Notes in Computer Science (LNCS)''. pp. 343-352, 2010.