Design optimization is the process of finding the best design parameters that satisfy project requirements. Engineers typically use design of experiments (DOE), statistics, and optimization techniques to evaluate tradeoffs and determine the best design.

In typical design problems, there are often many design parameters to consider. Some design parameters may have a nonlinear effect on the performance metrics. Others may only take on discrete values. There are often multiple competing requirements and objectives to meet. A manual approach of adjusting one parameter at time tends to lead to suboptimal results. On the other hand, it may be too time-consuming to evaluate all possible options in the design space.

Design optimization addresses these challenges by leveraging numerical optimization techniques to automatically find optimal solutions while satisfying constraints. The process searches the design space more intelligently than brute-force sweeps of the design space. The iterative process of modifying the design is automated, reducing turnaround time and human error. Engineers also use statistical methods to explore sensitivities and understand the design space before running the optimization, as well as afterward to evaluate the robustness of optimal solutions.

MATLAB^® and Simulink^® provide a range of design optimization capabilities, including general tools for optimizing any kind of model, as well as more targeted tools for specific applications:

• Optimize single and multiple design objectives with Optimization Toolbox™ and Global Optimization Toolbox. Different optimization solvers are available to address challenges such as nonlinearity, multiple optima, discrete design choices, and expensive simulations.
• Tune design parameters in a Simulink model to meet objectives such as improved system performance and minimized energy consumption with Simulink Design Optimization™. Using design optimization techniques, you can meet both time-domain and frequency-domain constraints such as overshoot and phase margin. You can jointly optimize physical plant parameters, controller gains, or any design parameters in your model to maximize overall system performance.
• Perform design of experiments to specify test plans, generate random numbers for Monte Carlo simulations, use sensitivity analysis to determine the robustness of your results, and create response surface models with Statistics and Machine Learning Toolbox™. For sensitivity analysis with Simulink models, use Simulink Design Optimization.
• Define test plans, develop statistical models, and generate optimal calibrations and lookup tables for complex powertrain systems with Model-Based Calibration Toolbox™.