A hybrid sequence sampling technique and its application to multi-objective optimization of blending process

Sampling technique is a method that exacts a certain number of samples from the overall sample set, and could be used to solve multi-objective optimization problem. This paper at first analyses the advantages and disadvantages of the three sampling techniques that Monte Carlo Sampling(MCS), Orthogonal Latin Hypercube Sampling(OLHS) and Hammersley sequence Sampling(HSS), then puts forward a hybrid sampling technique(HST) to make full use of the good 1-dimensional uniformity in OLHS and the good multidimensional uniformity in HSS. The advantage of the HST in the uniformity of the sample space is illustrated with comparison to the former three sampling techniques. Based on the HST, a method to solve multi-objective optimization problem is described. It selects every single objective of the problem in turn and converts the other objectives to inequality to form a single objective optimization, finally solves every single objective optimization problem to generate the Pareto set of the original multi-objective optimization problem. The method is applied to the multi-objective optimization of blending process in alumina production. Application case illustrates the feasibility and effectiveness of the proposed sampling technique and the optimization method.