Design Report Guided Missile Submarine SSG(X) - AOE - Virginia ...

SSG(X) Design – VT Team 3 Page 45
is then determined through summing SWBS labor costs, production support labor cost, and design and
integration labor cost.
• The material cost is determined using the SWBS weights, material cost factors, inflation factor, battery
type, propulsion propeller type, and manning and automation factor.
• The total direct cost is determined by summing the total labor cost with the total material cost.
• The total indirect cost is determined by multiplying the total direct cost with the overhead rate.
• The BCC is determined by summing the total direct and indirect costs and multiplying by one plus the
profit margin.
3.5 Multi-Objective Optimization
The Multi-Objective Optimizer uses a genetic algorithm to find the most effective feasible designs while
minimizing cost and risk and maximizing effectiveness. The Multi-Objective Genetic Optimizer (MOGO) chooses
a random population from this design space. Each individual design of the population is defined by its values of the
design variables. The synthesis model evaluates the effectiveness, cost, and risk of all individual designs in the
population. It determines the feasibility of each design and the fitness-dominance layers. After evaluation of the
initial population, the MOGO performs selection crossover mutation to define a new population. This new
population is influenced by the designs in the former population that performed favorably. Figure 33 shows a flow
chart of the MOGO process.
Figure 33 - Multi-Objective Genetic Optimization (MOGO)
The MOGO is implemented using the Darwin optimizer and it is integrated into Model Center along with the
synthesis model. The objectives, design variables, and constraints must be identified before use of the MOGO. The
objectives are to maximize the effectiveness (OMOE) and minimize the cost (CBCC) and risk (OMOR). The
feasibility ratios are the constraints and are given lower and upper bounds. The lower bound for the feasibility ratios
is set to zero. The design variables are added from the Input Module and lower and upper bounds are set for the
continuous variables.
The MOGO is added to the synthesis model as a new component. Before the optimization is run, the
parameters are adjusted so the population size is 200 with 60 preserved designs and the maximum number of
generations is 1,000. Evaluation of 100 populations without improvement defines convergence. For discrete
variables, the crossover probability is 1 and the mutation probability is 0.15; for continuous variables, the crossover
probability is 1 and the mutation probability is 0.1. The maximum constraint violation is 0.02 with a percent penalty
of 0.5 for violation.
The result of the MOGO is the non-dominated frontier; these designs are known as the Pareto designs. Figure
34 shows a three-dimensional representation of the non-dominated frontier.
3.6 Optimization Results
Figure 34 shows the non-dominated frontier from the optimization results and the chosen design, run number
44. The selected design has a cost of $633 million and an OMOE of 0.896. All non-dominated designs have the
highest effectiveness for a given cost and risk. Figure 34 is a 2-D representation of the Non-Dominated Frontier