Genetic algorithms(GAs) are general-purpose global optimization algorithms based on natural evolution principle. Real-coded genetic algorithms(RGAs) are concerned for its high precision and searching ability in whole space. This paper studies the efficiency of linear crossover operator in RGA by both mathematical analysis and simulation. A new linear crossover operator with optimum distribution is proposed, by which the son-individuals can reach uniform probability distribution in whole searching space. The application of the optimum distributed RGA in on-line optimization of a propylene hydration reactor is satisfied. Copyright 1999 IFAC