Instrumentation design and upgrade for principal components analysis monitoring

In this paper, a methodology to design instrument networks optimally selected for the use of PCA-based monitoring in process plants is presented. Sensors are placed or signals are selected to minimize the sum of investment cost and the cost of not detecting a fault. This optimization problem is solved using Genetic Algorithms (GA). It is shown that for existing sensor networks, the use of fewer signals than the available ones can be, in some cases, advantageous for monitoring purposes.