To date, researches on agent multi-issue negotiation are mostly based on linear utility functions. However, the relationship between utilities and resources is usually saturated nonlinear. To this end, we expand linear utility functions to nonlinear cases according to the law of diminishing marginal utility. Furthermore, we propose a negotiation model on multiple divisible resources with two phases to realize Pareto optimal results. The computational complexity of the proposed algorithm is polynomial order. Experimental results show that the optimized efficiency of the proposed algorithm is distinctly higher than prior work.