Using CFD and machine learning to explore chaotic mixing in laminar diameter-transformed stirred tanks

Elimination of segregated zones is important to intensify the global chaotic mixing of high-viscosity fluids. Diameter-transformed stirred tank (DTST) reactor as a new process-intensification equipment is proposed. Through the PLIF technique, tracer transport CFD modeling, and Poincare sections, we investigate the mixing structures inside a DTST reactor, and find that the DTST reactor can trigger global chaotic mixing. This is due to the wall squeezing effect, causing the fluid particle in periodic motion to become disordered. From the perspective of mixing kinetics, an area coverage model A = A(max)(1 - e(-Eamt)) is developed to quantitatively characterize its mixing efficiency, and the maximum area coverage Amax is 98.12 %. In addition, to help practitioners quickly select a suitable DTST reactor under different working conditions, machine learning-based flow pattern prediction models are proposed, among the ANNc model has a highest accuracy of 97.6 %. The decision boundary can be expressed by the Rec-E discriminant lg(Re-c) = 2.3/(E - 0.95)(0.18) - 1.