5.9 Selecting a Baseline and CCD Formulation

Two separate choices¶

A rigorous study makes two decisions:

Which non-CCD baseline represents current practice?
Which CCD formulation is appropriate for the coupled problem?

Use one-pass sequential design when it accurately represents the existing disciplinary handoff. Use an iterated sequential baseline when repeated engineering negotiation is the relevant practice. Neither should be presented as a CCD method.

Choose nested CCD when a reliable control solver already exists, elimination greatly reduces the outer dimension, and inner solutions can be converged and differentiated. Choose simultaneous CCD when coupling is strong, states and controls are naturally transcribed, sparse derivatives are available, or repeated inner solves are unreliable.

Decision evidence¶

The selection should be based on problem structure, not only the lowest runtime from one initialization. Report feasibility, objective, derivative quality, memory, restarts, and validation error.

Activity 5.9: select the baseline and CCD formulation¶

Activity 5.9: select the baseline and CCD formulation

Option	Classification	Validated $J$	CPU	Memory	failure rate	max constraint
One-pass	baseline	0.91	1.0 h	2 GB	0%	0.97
Iterated	baseline	0.84	3.5 h	3 GB	10%	1.02
Nested	CCD	0.73	12 h	6 GB	5%	0.99
Simultaneous	CCD	0.71	7 h	28 GB	20%	0.98

The available machine has 16 GB RAM. Require max constraint $≤1$ and failure rate no greater than 10 percent.

Select the fair baseline representing the simplest current workflow.
Identify feasible CCD formulations under the hardware and reliability limits.
Compute validated improvement of the selected CCD method relative to the baseline.
If a 32 GB node becomes available and the simultaneous failure rate drops to 8%, recompute the CCD choice.
Explain why the iterated option is not reclassified as CCD even if its objective improves.