Volume 1, Issue 1
Ordinal-Optimization-Based Framework for Optimization of Cooling Conditions for Reducing Distortion in Hot-Rolled Asymmetric Sections
Modeling and simulation tools are used increasingly to model thermo-mechanical processing of materials. Often these simulations are computationally expensive, and use of formal optimization tools to enhance product or process performance, or both, is practically infeasible in view of the large search space (variables), leading to a large number of function evaluations. Ordinal optimization (OO) adopts a different strategy compared to traditional optimization algorithms. It uses the “order” in the performances among designs, rather than the “value” and provides set of “good enough” solutions with a guarantee, instead of a unique best solution. It has been successfully applied to optimization of such computationally intensive problems. OO can be easily integrated with other optimization algorithms. OO’s integration with genetic algorithms (GA) leads to a hybrid algorithm called “genetic ordinal optimization” (GOO). It has better stopping criteria, along with a high probability of finding truly “good enough” design compared to the traditional GA. Cooling of asymmetric sections, post hot rolling, leads to a large distortion because of thermal gradients and phase transformations. Lengthwise distortions pose operational difficulties and challenges for post-rolling operations, including straightening. Here, a possible scheme of a sequence of forced convection cooling stages is considered in place of the traditionally used free convection on the cooling bed. In this work OO, GOO, and GA are applied to optimize the sequence of forced cooling for minimal distortion. A simplified regression model for distortion, built using results of a detailed thermo-mechanical simulation model, is used in the optimization loop. Results obtained show reduction in the distortion by a significant amount as compared to the natural convection cooling on the cooling bed. Besides providing a possible solution, this work also shows the efficacy of OO and GOO for application to industrial problems.