Trans RINA, Vol 154, Part C1, Intl J Marine Design, Jan - Jun 2012


certainty; the measure gives some indication of the scope of VCG change that can be accommodated whist still passing the criteria. Damage stability was not considered since the vessel compartmentation would not yet have been defined at this initial design stage.


2.4 ANALYSIS AUTOMATION


To complete the design-space exploration, an automated method of producing design variants and submitting them for analysis as well as results collation and post- processing


is required. In the current work the


FRIENDSHIP-Framework (FFW) was used to define the parametric model as well as the other tasks listed above. The FFW and the simulation software are developed by different software vendors. However, all the software used has been designed so as to allow inter-process communication thus enabling the FFW to manage the overall analysis. Suitable integration mechanisms were developed in the FFW to export the hull geometry and then import


this geometry and run the analyses in


SHIPFLOW, Hydromax (HM) and Seakeeper (SK). The results of the analyses were then read back into the FFW for post-processing to calculate the final performance measures for each variant. Figure 5 shows a typical screenshot of HM and SK being driven by the FFW.


the analysis in a reasonable timeframe. It is generally not feasible to explore the design-space with a regular mesh- like variation of the parameters; the number of variants would be the number of different values used for each parameter raised to the


power of the number of


parameters: for the current example, four parameter values for each of the nine parameters would generate 49 –over 262,000– variants! For this reason, a “Design of Experiments” (DoE) approach is sensible for efficient and economical


coverage of the design-space. This


approach changes all the design parameters at the same time between different variants, rather than just one single parameter at a time. Response surfaces are then used to allow interpolation of the performance measures between the discrete values calculated for the individual variants.


2.5 (a) Design of Experiments


To generate the variants, a Sobol algorithm [11] is used to yield a quasi-random, yet uniform sampling of the values of the design parameters over the desired range (Table 2). Two-hundred variants were generated (analyses of which would take a couple of days on typical desk-top hardware). The Sobol algorithm has the advantage of being repeatable (rather than a true random process) although it still exhibits random-like scatter. This means


that it is easy to run each analysis


independently (each on a different computer for speed, for instance) or to run subsequent analyses (should a new performance measure be required) whilst ensuring that exactly the same design variants are used each time.


2.5 (b) Response Surfaces


The design-space exploration generates a large quantity of data and represents not insignificant computational effort (especially if sophisticated numerical simulation tools have been used). It is useful then to reuse this data, potentially for automated optimisation or other similar applications. Following the work of


Harries [4], a


response surface, meta-model method has been used to capture the simulation data and allow interpolation of the performance measures without


having to redo the


numerical simulations. In the current work, response surface models (RSM) are fitted to the performance data using a Kriging approach [12].


Figure 5: Screenshot of FRIENDSHIP-Framework,


Hydromax and Seakeeper in use. 2.5 DESIGN-SPACE EXPLORATION


Even for relatively simple cases, the design-space can have many dimensions (for the example presented, nine parameters are used to define the model, hence a nine- dimensional design-space). For this reason, efficient exploration of the design-space is required to complete


©2012: The Royal Institution of Naval Architects


The Sobold algorithm populates the design-space efficiently, but it does so by varying all the design parameters at the same time between design variants. (So it its not possible to see the effect of varying just a single parameter, with all other parameters remaining fixed, directly


from the raw data of keeping all the


exploration.) The RSM provide a means for smooth interpolation between these discrete data (for example varying length while


other parameters


constant). It should be noted that since the exploration algorithm does not generate variants that cover the entire design-space, great care should be taken to ensure that


design-space


C-21


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57  |  Page 58  |  Page 59  |  Page 60  |  Page 61  |  Page 62  |  Page 63  |  Page 64