(268h) Efficient Breadth-First Tree Search Method for Nonlinear Sensor Network Design

Sensor networks are important tools to gather information to properly monitor the plant. The problem of systematically locating sensors that meets pre-specified criteria, formally known as the sensor network design problem, is of interest for a past few decades. Recently, middle scale and large scale sensor network design problems were solved using tree enumeration methods that are based on characteristic sets of the process systems, namely cutsets in linear systems (Mayur and Bagajewicz, 2006a, 2006b) and balance equations in nonlinear systems (Nguyen and Bagajewicz, 2008). However, realistic design problems for large scale nonlinear systems were not solved efficiently by these methods.

In this work, we propose two special breadth-first tree search methods with efficient pruning strategy for solving realistic large scale nonlinear design problems. The proposed tree search methods are based on individual measurements instead of characteristic sets. The methods guarantee optimality and are shown to be satisfactorily efficient. A middle scale example and a large scale example, the Tennessee Eastman problem, will be used to illustrate the power of the new method. We will show orders of magnitude reduction in computational time

References

Gala, M. and Bagajewicz, M. J. Rigorous Methodology for the Design and Upgrade of Sensor Networks Using Cutsets. Ind. Eng. Chem. Res. 2006a, 45(20), 6687-6697.

Gala, M. and Bagajewicz, M. J. Efficient Procedure for the Design and Upgrade of Sensor Networks Using Cutsets and Rigorous Decomposition. Ind. Eng. Chem. Res. 2006b; 45(20), 6679-6686.

D. Nguyen and Bagajewicz M. Design of Nonlinear Sensor Networks for Process Plants. Industrial and Engineering Chemistry Research. 2008, 47(15), 5529-5542