Double surrogate modeling usage in PSO

Purpose The purpose of this paper is to improve the efficiency of particle optimization method by using direct and indirect surrogate modeling in inverse design problems. Design/methodology/approach The new algorithm emphasizes the use of a direct and an indirect design prediction based on local surrogate models in particle swarm optimization (PSO) algorithm. Local response surface approximations are constructed by using radial basis neural networks. The principal role of surrogate models is to answer the question of which individuals should be placed into the next swarm. Therefore, the main purpose of surrogate models is to predict new design points instead of estimating the objective function values. To demonstrate its merits, the new approach and six comparative algorithms were applied to two different test cases including surface fitting of a geographical terrain and an inverse design of a wing, the averaged best-individual fitness values of the algorithms were recorded for a fair comparison. Findings The new algorithm provides more than 60 per cent reduction in the required generations as compared with comparative algorithms. Research limitations/implications The comparative study was carried out only for two different test cases. It is possible to extend test cases for different problems. Practical implications The proposed algorithm can be applied to different inverse design problems. Originality/value The study presents extra ordinary application of double surrogate modeling usage in PSO for inverse design problems.

Determination of the Optimal Locations for Injection Molding Gates with Higher Order Response Surface Approximations

This paper proposed a gate location optimization scheme to minimize the maximum injection pressure in plastic injection molding. The method utilized a series of higher order response surface approximations (RSA) to model the maximum injection pressure distribution with respect to gate locations, and the global minimum of these response surface models were subsequently sought by a global optimization method based on a multi-start sequential quadratic programming technique. The design points for RSA were evaluated by the finite element method. After a sequence of repetitions of RSA and optimization, the converged minimizer would represent the optimal gate location. A rectangular plate with two segments of different thicknesses was selected to demonstrate the effectiveness of the procedure. The variation of the thicknesses causes the optimal gate location to deviate from the center and induce multiple valleys in the maximum injection pressure distribution, which is ideal for the application of the higher order RSA and a global searching technique.

Optimized Multi-Floor Throughflow Micro Heat Exchangers

Multi-floor networks of straight-through liquid cooled microchannels have been investigated by performing conjugate heat transfer in a silicon substrate of size 15×15×1 mm. Two-floor and three-floor cooling configurations were analyzed with different numbers of microchannels on each floor, different diameters of the channels, and different clustering among the floors. Thickness of substrate was calculated based on number of floors, diameter of floors and vertical clustering. Direction of microchannels on each floor changes by 90 degrees from the previous floor. Direction of flow in each microchannel is opposite of the flow direction in its neighbor channels. Conjugate heat transfer analysis was performed by developing a software package which uses quasi-1D thermo-fluid analysis and a 3D steady heat conduction analysis. These two solvers are coupled through their common boundaries representing surfaces of the cooling microchannels. Using quasi-1D solver significantly decreases overall computing time and its results are in good agreement with 3D Navier-Stokes equations solver for these types of application. Multi-objective optimization with modeFRONTIER software was performed using response surface approximations and genetic algorithm. Maximizing total amount of heat removed, minimizing coolant pressure drop, minimizing maximum temperature on the hot surface, and minimizing non-uniformity of temperature on the hot surface were four simultaneous objectives of the optimization. Pareto-optimal solutions demonstrate that thermal loads of 800 W cm−2 can be effectively managed with such multi-floor microchannel cooling networks. Two-floor microchannel configuration was also simulated with 1,000 W cm−2 uniform thermal load and shown to be feasible.