Enhancement of Mixing Performance of Two-Layer Crossing Micromixer through Surrogate-Based Optimization

Abstract

Optimum configuration of a micromixer with two-layer crossing microstructure was performed using mixing analysis, surrogate modeling, along with an optimization algorithm. Mixing performance was used to determine the optimum designs at Reynolds number 40. A surrogate modeling method based on a radial basis neural network (RBNN) was used to approximate the value of the objective function. The optimization study was carried out with three design variables; viz., the ratio of the main channel thickness to the pitch length (H/PI), the ratio of the thickness of the diagonal channel to the pitch length (W/PI), and the ratio of the depth of the channel to the pitch length (d/PI). Through a primary parametric study, the design space was constrained. The design points surrounded by the design constraints were chosen using a well-known technique called Latin hypercube sampling (LHS). The optimal design confirmed a 32.0% enhancement of the mixing index as compared to the reference design.