In recent years

immune genetic algorithm (IGA) is gaining popularity for finding the optimal solution for non-linear optimization problems in many engineering applications.However

IGA with deterministic mutation factor suffers from the problem of premature convergence.In this study

a modified self-adaptive immune genetic algorithm (MSIGA) with two memory bases

in which immune concepts are applied to determine the mutation parameters

is proposed to improve the searching ability of the algorithm and maintain population diversity.Performance comparisons with other well-known population-based iterative algorithms show that the proposed method converges quickly to the global optimum and overcomes premature problem.This algorithm is applied to optimize a feed forward neural network to measure the content of products in the combustion side reaction of

p

-xylene oxidation

and satisfactory results are obtained.