Both Gradual Kriging Optimisation tool GKO and PRObabilistic MultI CritEria design system PROMICE are INASCO's current achievements in the field of Design Optimisation

Slide 1

PRObabilistic MultI CritEria design system (PROMICE) is an in-house optimisation tool. It maximizes the Probability of Success, POS of multiple design criteria simultaneously by taking into account uncertainties arising  either from the investigated system or the structure.
 jpdm3_small.png During PROMICE optimisation, each design alternative is represented in a multidimensional space as a multivariate statistical distribution. Each dimension represents design criteria (drag and weigth in this example) with different allowable values, and optimisation is performed by maximising the probability of satisfying both criteria (Joint Probability of Success) simultaneously. In this example, three design alternatives are evaluated in terms of Drag and Weight.
PROMICE can be applied on every Design phase (Conceptual, Preliminary, Detailed) as long as system models and uncertainty information are available in any format. It is currently used in MAAXIMUS project as a core module of the Probabilistic Life - Cycle Management platform and in NACRE project in three different tasks.
INASCO involvement in NACRE: Evaluation of Cabin – Structural – Aerodynamic concepts, Engine positioning optimization and Manufacture-driven wing optimization. 
Gradual Kriging Optimisation (GKO) is a Matlab - based MDO (Multidisciplinary Design Optimisation) tool which includes sophisticated methods such as Kriging and Optimal Latin Hypercube. Its philosophy includes the sequential generation of Kriging metamodels in a gradually reduced design space until it reaches optimum solution with accepted accuracy.
GKO graphical user interface module for the generation of optimal Design of Experiments

Kriging is a statistical interpolation method used in MDO problems to replace computationally expensive models, and it is mainly used for approximating systems with a large amount of variables using a relatively low amount of high – fidelity information (FEM or CFD calls). Optimal Latin Hypercube generates a maximum quality sampling in a multi – dimensional Design Space in order to build a robust Kriging Approximation. (Ref: HISAC , MUSCA project)
Slide 3
Output graphs of GKO tool: Optimal Latin Hypercube (left) creates a low - discrepancy distribution of sample points. This is the first step to generate an accurate and numerically stable Kriging (right, coloured surface) approximation model which replaces any "expensive" analysis model (white surface).