Multiobjective optimization of mechanical properties based on the composition of adhesives

Abstract

A mixed numerical-experimental approach capable to predict and optimize the performance of the footwear adhesive joints, based on the weight composition of used raw materials was presented. The approach based on the optimal design of adhesive composition to achieve the targets of minimum creep rate (CR) and maximum peel strength (PS) under manufacturing. Two stages are considered in the proposed approach. In the first stage, an approximation model is built based on planned experimental measurements and artificial neural network (ANN) developments. The ANN learning procedure uses a genetic algorithm. In the second stage an optimal design procedure is developed based on multi-objective design optimization (MDO) concepts. The MDO algorithm based on dominance concepts and evolutionary search is proposed aiming to build the optimal Pareto front. The model uses the optimal ANN to evaluate the fitness functions of the optimization problem. Furthermore, a ANN-based Monte Carlo simulation procedure is implemented and the sensitivity of the CR and PS relatively to weight compositions of raw materials is determined. The approach shown robustness to establish the trade-off between minimum CR properties and minimum inverse PS (maximum PS) using the weight composition of used raw materials. The optimal results for both CR and PS based on proposed approach are reached when large quantities for polyurethanes (Pus) and for some additives are considered. The performances of adhesive joints measured by CR and PS are very sensitive to the influence of some PUs and in some way are moderately sensitive to additives. The proposed MDO approach supported by experimental tests shows improved explorative properties of raw materials and can be a powerfully tool for the designers of adhesive joints in footwear industry.