A Genetic Algorithm-Based Optimization Approach for Fixture Design that Preserves Desired Dynamic Response Characteristics

Aerospace vehicles are complex systems comprised of interconnected components that must perform reliably together as an assembly. However, reliability tests are often performed at the component level, which can lead to poor results if the interconnections between components are not accurately represented during testing. In the ideal case, a system’s components would be tested in situ such that the boundary conditions of the components during testing would be the same as in operation. However, having access to the connecting components of a part can be difficult due to barriers such as inaccessible proprietary design specifications and different design timelines. Therefore, there is a need to simulate the boundary conditions provided by a connecting component without having direct access to the component. In this chapter, a process for designing a test fixture that accurately simulates the boundary conditions of a connecting component is proposed. It is assumed that the response of the connecting component is known, and the goal of the process is to design a fixture that replicates that response, without specific knowledge about the original component’s geometric dimensions or material properties. The design is determined using a genetic algorithm-based optimization process that chooses combinations of bounded design parameters until the response of the fixture matches the desired response. The optimization starts by subdividing a design volume, where each partition can be assigned a different density and Young’s modulus value. At each iteration, a different combination of parameter values, determined by a genetic algorithm, is chosen. The natural frequencies and mode shapes of the current design are determined and compared to the target values. Once a sufficient match is achieved, the design is complete. To demonstrate the viability of the proposed approach, a simple component was designed and modeled. The component’s first five natural frequencies and mode shapes were determined and used as the basis for the objective function. The optimization algorithm was run and a component design was determined without assuming any a priori knowledge of the target component’s material properties or geometry, with the exception of its overall dimensions. The resulting component’s first five natural frequencies were within \(16\%\) of the target values and the Modal Assurance Criterion values for the first five mode shapes were all greater than \(0.79\) .