A hybrid method combining a genetic algorithm with a flexible tolerance algorithm is proposed for global optimization problems with multiple nonlinear constraints and peaks. The adaptive genetic algorithm is used to localize the “best” areas

while the flexible tolerance algorithm exploits this area by search mechanism for quasi-feasible point. To evaluate the efficiency of this method

a complex function with six peaks and four constraints is implemented and compared with the results supplied by sequential uniconstrained minimization technique(SUMT)

which indicates that the hybrid method is able to improve convergence and reduce computing task greatly.