Multi-step evolution method in genetic algorithms based on contract precision

To effectively overcome the disadvantages of traditional genetic algorithm whose terminate rule was designed subjectively and enhance the search efficiency during the evolution period, the definition of contract precision was presented according to the feedback control principle, and then the evolution period was divided into three parts based on contract precision. To enlarge the difference in individuals, different fitness methods were adopted in different phases. And the cross operator was improved with auto-adaptive crossover probability and uncorrelated matching to realize a higher convergence speed. The calculation results show that the improved GA is more scientific and efficient.