b'Engineer Innovation | ManufacturingThe challenge for the SANDEN engineers, then, was to speed up the discovery of duct shapes that met all their performance criteria. Their initial approach was to automate the traditional design of experiments (DOE) plan they would have performed manually in past years to determine the effects of changing a small number of design variables. Using a Java macro, they developed a procedure to automate the evaluation of design points, but found that it was slow and cumbersome, even when considering as few as five design parameters. They then turned to Figure 4: The 12 design parameters used in the Simcenter STAR-CCM+ Design Manager study. Design Managerthe design exploration tool within Simcenter STAR-CCM+, which is purpose-built for conducting automated exploration of the design space-to guide their design process.As illustrated in Figure 3, design exploration with Simcenter STAR-CCM+ represents a fundamentally new way of conducting engineering design. The traditional approach (such as the DOE plan mentioned earlier) involves simply simulating the CAD model, assessing whether the performance of a particular design is good enough, then performing successive simulations on variants of that design until performance criteria are met. This was Figure 5: Comparison of baseline design and result of Design Manager exploration study. a perfectly reasonable approach back when computing resources were limited, but it suffers from a number of significant drawbacks, including Baseline values Improved designPercentage variation limited model fidelity and the inability values to handle large numbers of design A R19 R20 5.3% parameters. True design exploration B 0,1 0,10% starts by defining the performance criteria (the objectives) upfront, and C 5 7,3 46% then using intelligent search D 419 mm 390,5 mm -7.0% algorithms (or better yet, multiple E 6 mm 0,1 mm -98.3 % algorithms) to discover families of F 5 mm 12 mm 58.30% better designs. This allows for using G 3 4 -33.3% computational models of high fidelity (to capture the pertinent physics), H R30 R48 mm 37.50% parameterized using a large number of I 91,816 m 90 mm -2.02% design variables. J R10 R100%K R20 R39 48.70% In the case of the ducting for the L R20 R15 -33.30% SANDEN evaporator, the engineers used 12 design parameters in their Figure 6: Comparison of design parameter values of baseline design and DesignDesign Manager study, as shown in Manager improved design. Figure 4.42'