Due to economic reason, not every process variable can be measured by a sensor. The problem of optimum selection of sensor location is referred to as sensor network design problem (SNDP).

Being a combinatorial optimization problem, the SNDP poses significant computational challenges for researchers, especially for large scale problems. The methods to solve the SNDP can be divided into two three classes: mathematical programming, graph-theoretic methods and stochastic methods (e.g. genetic algorithm)

The SNDP problem itself can be divided into two big classes: designing sensor network intended for process monitoring purpose (to obtain accurate process data) and designing sensor network for process fault diagnosis and resolution. The former can be solved by many methods while the latter is usually solved by graph-theoretic methods

Although extensive researches have been done on this problem, efficient methods to design sensor networks for large scale nonlinear problems have not yet been found. Moreover, all the published models are developed from technical point of view, which requires knowledge / expertise of the users to use appropriate constraints / specifications in the model. A model that bases solely on an economic viewpoint has not yet been proposed.

Addressing the mentioned drawbacks is the objective of this work. More specifically, in this work:

i)Efficient computational methods to solve SNDP for large scale nonlinear problems are proposed.

ii)A value-optimal SNDP is proposed and solved by using appropriate methods.