2.3    Optimiser Objectives

Optimisation Objectives can be configured in the Objectives tab of Configure Optimiser dialogue and are displayed in the Objectives list.

See Objectives chapter for more information.

The Goal Function calculated during the optimisation routine is based on the defined Objectives in the following way:

Consider an Objective defined in a Sub-Range of N frequency / angle points. Let O[i], i=1, 2…N, denote the current value of this Objective at the ith point in a Sub-Range. Similarly, let T(i), UB(i) and LB(i) denote the values of Target, Upper and Lower at the ith point. Note that in most cases, T(i)=const(i), UB(i) =const(i) and LB(i)= const(i); the values of Target, Upper and Lower vary within the Sub-range only in the case of Objective from file.

The value of the Goal Function G for the above Objective is:

if UB is defined, define GU:GU = Max (O[i]-UB[i])/(UB[i]-T[i]), i=1, 2…N,

if LB is defined, define GL: GL = Max (LB[i]-O[i])/(T[i]-LB[i]), i=1, 2…N,

if UB is defined and LB is not:G = GU

if LB is defined and UB is not:G = GL

if both LB and UB are defined:G = Max (GL, GU)

When there are several Objectives in the optimisation project, then the maximum value of the individual goal function G for different Objectives is taken as the value of the global Goal Function for the project, at this particular optimisation step.

With the above definitions, when the Goal Function drops below zero, it means that all the desired bounds (upper and lower) for all Objectives have been reached. That is why Goal Function equal zero is a natural stop criterion for the QW-OptimiserPlus.

Formally, the Add Objective or Edit Objective dialogue allows defining only target T for an objective, without defining either UB or LB. Then, the Goal Function G for such an objective is calculated as:

G = Max (Abs(O-T))

However, operation without bounds, with target only, is not recommended. Note that it does not distinguish whether the current result is below or above target.

The reasoning should be as follows: our actual goal is to push the result above the lower bound LB and/or below the upper bound UB. The lower and /or upper bounds define our limits of an acceptable design. The value set up as target T may be rather “idealistic”, and may not even be feasible in the practical design. The key role of target T in our optimisation project is to change the scaling of the Goal Function for a particular Objective. Thus modification of the Target value can be used to set a proper balance between the weights of different objectives.