Pareto set

Table 18.8 Comparison of Objective Functions and Decision Variables for a few Chromosomes in the Pareto Sets shown in Figrue 18.48...

Rudolph, G. (1998). On a multi-objective evolutionary algorithm and its convergence to the Pareto set, in Proceedings of the 5th IEEE Conference on Evolutionary Computation (IEEE Press, Piscataway, New Jersey), pp. 511-516. 

Fig. 4.1 The Pareto set obtained for the ZDT4 problem (Deb, 2001) using NSGA-II-JG. An additional point, C, is also indicated...

The procedure used in SSA has been extended to multi-objective problems by Suppapitnarm et al. (2000). These workers used the neighborhood perturbation method of Yao et al. (1999) to create a new point around an old point. This algorithm is known as multi-objective simulated annealing (MOSA). Since a Pareto set of solutions is to be... 

Fig. 4.12 Plots of non-dominated solutions obtained with NSGA-II-JG after 240,000 -300,000 function evaluations (fn. evals.) and for NSGA-II for 320,000 - 400,000 fn. evals. for the ZDT4 problem. Note that /j and I2 extend over [0, 1] (global Pareto set) for NSGA-II-JG only after about 300,000 function evaluations, and do not show this characteristic for NSGA-II...

Return to an uncrowded base point periodically (so as to generate a more continuous Pareto set, by locally exploring around uncrowded points in the non-dominated set)... 

Inverted Generational Distances for problems ZDTl, ZDT2 and ZDT3 are given in Table 5.1 for a better understanding. The true Pareto sets for ZDT test problems are represented by 100 solutions on the Pareto front. It is seen that IGD for the non-dominated solutions obtained by SAEA is smaller than that of NSGA-II. Other metrics can also be used to determine and compare results between different optimization algorithms. 

Das, I. and Dennis, J. E. (1997). A closer look at drawbacks of minimizing weighted sums of objectives for Pareto set generation in multicriteria optimization problems, Structural Optimization 14, 1, pp. 63-69. 

Viennet, R., Fonteix, C., and Marc, I. (1996). Multicriteria optimization using a genetic algorithm for determining a Pareto set. International. J. Syst. Sci., 27, 2, 255-260. 

Zhang, Y., et al. (2009) also used a laboratory SMB set-up to carried out three experimental runs to verify the optimization results corresponding to the points in the Pareto sets of case 2. It was found that for SMB as well as Varicol operation the recoveries and the extract purities are much lower then the optimization calculations. The authors explain that the major cause for such deviations stems from limitations of the pump performance and hence severe fluctuations of the recycling flow rate and point out that their experimental results indicate the great importance of flow rate control and pump performance in SMB and Varicol operation . Therefore, they also recommend that operating conditions for SMB and Varicol processes should be selected as a compromise between separation performance and robustness . 

Each point in the Pareto curve represents an optimal combination of the objective functions. The selection of the operating conditions for a production is a matter of additional economic considerations. Any point above the Pareto curve has tvorse performance than the points constituting the Pareto curve, because there is aWays one element of the Pareto set which has a lower solvent consumption for the same productivity. The points located below the Pareto curve are not feasible because the purity requirements are not fulfilled. ... 

To assess the performance of the proposed procedure for the Pareto-set of the example (see Figure j.lS), the goodness of approximation is tested by comparing the desired performance vector y with the performance of the configuration determined with the described procedure (H x)). Ideally, both performance vectors are equal. To measure the. similarity of both performance vectors the relative deviation off w.r.t. its corresponding. simulated performance A x) is calculated by Bj = I contains the relative deviation of... 

Figure 4 shows the Pareto set approximation found. In the Figure, the first extreme point located at (0,0) represents the solution where no link is damaged. The point located at (22.2 %, 1) corresponds the maximum importance in the network when considering the damage of a single link (link 3). 

Figure 6 and 7show the Pareto set approximation found using the MO-PSDA implementation. Out of the total 2 potential solutions for this problem, the Pareto set was identified by analyzing only a total of 5000 combinations. 

It is important to observe that Pareto sets usually define a region of operating variables however, a single operation policy should still be selected for real implementation. Therefore, the user still has to rely on some sort of arbitrary procedure to select the best operation policy among the possibly infinite number of solutions that constitute the Pareto set. Besides, solutions in the Pareto set can indeed lead to very poor reactor operation, as none of the objectives is reached optimally. 

Multi-objective optimization procedures were used for the simultaneous maximization of monomer conversion and minimization of side products during low-density polyethylene polymerizations performed in tubular reactors under steady-state conditions [170]. Genetic algorithms were used to compute the Pareto sets. Multi-objective optimization procedures were also used for the simultaneous maximization of molecular weight averages and minimization of batch times in epoxy semibatch polymerizations [171]. In this case, monomer feed rates were used as the manipulated variable. 

During the last twenty years, the literature on MCDM problems has grown at a high rate where few techniques for generating the Pareto optimal have been well developed and evaluated. The idea of the Pareto optimality and generating the Pareto set, are briefly reviewed here. These concepts and techniques are dealt with in more detail by Ref. [17-22]. 

The entire Pareto set can be generated by repeatedly solving the above problem for different sets of constraint levels Si. The computational effort is the main disadvantage of this technique as the number of constraints increases and the simplified optimisation problem is required to be solved several times. 

The Pareto set, i.e., the set of aU nondominated solutions of an /M objective problem can be constructed by assembling all solutions x, of all aggregated optimization problems of the form... 

---