DML-IDE (Design Model Language IDE) is a topology optimization program built off the massively parallel physics engine Titan. Titan uses CUDA to simulate the behavior of large mass-spring lattices, reaching up to 3 billion updates per second.

DML-IDE performs topology optimization by creating a random lattice within a given volume and simulating its response to load cases. The maximum stress in each bar is recorded, and the bars with the smallest stress are removed every optimization period. Then, more random masses and bars are added in a process called regeneration. Regeneration makes the final optimized structure of the parts very repeatable.

Optimization of a cantilever beam.

Frequency Optimization

For dynamic loads, optimization of the natural frequencies of the lattice structures can also be important for design. Following the simulation based philosophy of the weight reduction topology optimization mode, I developed a frequency optimization mode for DML-IDE.

A gradient-based frequency optimization method involves calculating the gradient of the natural frequencies with regards to the design variables. In the case of DML-IDE, the natural design variables are the position of each mass. The natural frequency gradient is a function of the natural frequency and the mode shape of oscillation. Both of these can be obtained via simulation only by calculating a Discrete Fourier Transform of the mass positions while the simulation is being run. This harnesses the full power of the parallel nature of Titan, and allows for weight optimization and frequency optimization to be performed simultaneously.