European LSDYNA Conference 2017
Our paper.
The oral presentation.
MY PRESENTATION AT SAE BRAKE IN SEPTEMBER
The oral presentation.
MY PRESENTATION AT ASME IN AUGUST
The oral presentation.
A RBDO BENCHMARK WITH 100 VARIABLES AND 80 CONSTRAINTS
I have continued to improve my SORMbased RBDO algorithm. Yesterday I think solved a rather challenging benchmark with 100 variables and 80 constraints.
The CPUtime was only 114 seconds and the accuracy extremely high.
THE BENCHMARK
This is a generalization of a benchmark with 10 variables
and 8 constraints suggested by Cho and Lee. It is formulated
for N times 10 variables and N times 8 constraints.
The problem is then solved using a new algorithm for
SORMbased RBDO with nonGaussian variables and the performance is excellent.
THE SOLUTION
In fact, the solution generated with this algorithm (SSQP) is slightly
better than presented previously ([14]) depending on the SORM
corrections utilized in the algorithm. The solution is obtained within 11 CPU seconds for 10 variables and 8 constraints.
THE PERFORMANCE
The performance of the algorithm is excellent. The benchmark with [2,3,4,5,6,7,8,9,10]
times 10 variables and [2,3,4,5,6,7,8,9,10] times 8 constraints are solved with high accuracy
and extremely fast.
The plot shows CPUtime and the optimal objective value as a
function of number of variables N. The CPUtime is 114 seconds for 100 variables and 80 constraints. In addition, the optimal objective
value is exactly 10 times the value obtained for 10 variables and 8 constraints.
