| Summary: | The goal of this research is to develop process monitoring technology capable of taking advantage of the large stores of data accumulating in modern chemical plants. There is demand for new techniques for the monitoring of non-linear topology and behavior, and this research presents a topological preservation method for process monitoring using Self Organizing Maps (SOM). The novel architecture presented adapts SOM to a full spectrum of process monitoring tasks including fault detection, fault identification, fault diagnosis, and soft sensing.
The key innovation of the new technique is its use of multiple SOM (MSOM) in the data modeling process as well as the use of a Gaussian Mixture Model (GMM) to model the probability density function of classes of data. For comparison, a linear process monitoring technique based on Principal Component Analysis (PCA) is also used to demonstrate the improvements SOM offers. Data for the computational experiments was generated using a simulation of the Tennessee Eastman process (TEP) created in Simulink by (Ricker 1996). Previous studies focus on step changes from normal operations, but this work adds operating regimes with time dependent dynamics not previously considered with a SOM.
Results show that MSOM improves upon both linear PCA as well as the standard SOM technique using one map for fault diagnosis, and also shows a superior ability to isolate which variables in the data are responsible for the faulty condition. With respect to soft sensing, SOM and MSOM modeled the compositions equally well, showing that no information was lost in dividing the map representation of process data. Future research will attempt to validate the technique on a real chemical process.
|
|---|
