Local stress constraints in topology optimization of structures subjected to arbitrary dynamic loads: a stress aggregation-free approach

We present an augmented Lagrangian-based approach for stress-constrained topology optimization of structures subjected to general dynamic loading. The approach renders structures that satisfy the stress constraints locally at every time step. To solve problems with a large number of stress constraints, we normalize the penalty term of the augmented Lagrangian function with respect to the total number of constraints (i.e., the number of elements in the mesh times the number of time steps). Moreover, we solve the stress-constrained problem effectively by penalizing constraints associated with high stress values more severely than those associated with low stress values. We integrate the equations of motion using the HHT-method and conduct the sensitivity analysis consistently with this method via the discretize-then-differentiate approach. We present several numerical examples that elucidate the effectiveness of the approach to solve dynamic, stress-constrained problems under several loading scenarios including loads that change in magnitude and/or direction and loads that change in position as a function of time.

Automated summary

The approach is based on an augmented Lagrangian method, effectively solving stress-constrained topology optimization problems for structures subjected to general dynamic loading. Normalizing the penalty term and penalizing constraints associated with high stress values more severely are key strategies employed for handling problems with a large number of stress constraints.