Optimization process

From Intamap

• Description of a scenario

The description of an optimization scenario is a combined description of the optimization problem and the solution taken.

• The optimization problem

An optimization scenario in monitoring network is related to a number of notions that lead to a full response of the automatic interpolation service for specific request from the client.

The first notion is that of target. The client defines a target related to the studied environmental variable. This target has certain properties that bring a gross classification. Main characteristics to consider are linearity and spatial support of interest (point-wise to global). The notion of target is a lot related to the situation faced by the client. for instance, in the radioactivity exposure context it is likely to have two types of situation: background and emergency. How these situation should be dealt with? For sure switching the situation will change the request from the client yielding different targets. Third network designs problems, and thus optimization scenarios, include constraints generally opposing the notions of quality and cost.

With a geographical sketch of the network and its evolution (deleting/adding new devices), the three notions of the above describe the optimization problem.

• The solution

Description of a chosen solution is related to three notions as well. First the approach corresponds to the chosen description of the environmental variable. If a suitable model is available (a linear model for instance) then model-based approaches are likely to be envisaged. Once the approach is chosen, the whole description of the optimization problem can then be translated into an optimization criterion. In the literature the notion of optimization criterion is closely linked to the optimization method/algorithm. Several options for the method may arise though when considering a defined criterion. For instance in the context of using a Universal kriging model, optimization of the mean universal kriging variance of the interpolated variable may be done with simulated annealing, a genetic algorithm or even combined methods.

• Flowchart