A novel genetic algorithm (GA) including chaotic variable named chaos-genetic algorithm (CGA) was proposed. Due to the nature of chaotic variable, i.e. pseudo-randomness, ergodicity and irregularity, the evolutional process of CGA makes the individuals of subgenerations distributed ergodically in the defined space and circumvents the premature of the individuals of subgenerations. The performance of CGA was demonstrated through two examples and compared with that by the traditional genetic algorithms (TGA). The results showed the superior performances of CGA over TGA, and moreover, the probability of finding the global optimal value by using CGA is larger than that by using TGA. To illustrate the performance of CGA further, it was employed to optimize the operating conditions of aromatic hydrocarbon isomerization (AHI) process modeled by the radial basis functions (RBF) coupled with partial least squares (PLS) approach. Satisfactory results were obtained. Further, a generalized methodology, which employs RBF-PLS approach to model the complex chemical process based upon practical observation data and subsequently applies CGA to find the optimal operating conditions, was suggested too. (C) 2003 Elsevier Science Ltd. All rights reserved.