It will greatly enhance productivity and save production cost if we properly use optimization algorithm in industrial production. Genetic algorithm is one of the most classical optimization algorithms. It is widely applied in the fields of function optimization, artificial intelligence, image processing, data mining and so on, because of its advantages, such as good inherent parallel-ability, high robustness, convenient operation and reliable results et al.. However, traditional genetic algorithm has disadvantages of slow convergent speed, low convergent precision and poor local search capability, which severely impact the actual application effect of genetic algorithm.To solve these problems, this paper proposes an adaptive genetic algorithm based on crossover library and parallel mutation by starting with crossover operator and mutation operators. Firstly, the shortage of seldom considering population situation in genetic algorithm is analyzed, then the population diversity based on eigenvalues, and the population diversity based on information entropy and individual distribution, are separately defined. To judge the effect of population evolution, the concept of the similarity between adjacent populations is also raised. Through considering the population diversity and the similarity between adjacent populations in an unified frame, population vigor, a new evaluation criterion of population is built. Using the population vigor, the crossover and mutation probabilities can be adaptively adjusted, where the reference value of population fitness is reset by mode fitness instead of average fitness, and both population vigor and individual fitness are fully considered when calculating the probabilities. To enhance the convergent speed, a dynamic crossover library is built to store individuals which well meet the evolutionary feature, where individuals will always be selected as parents to produce offspring in crossover operator. To improve the local search capability, mutation operator is improved by parallel mechanism and different mutation rules are simultaneously carried out in mutation operator. At last, the cargo transportation sruction optimization and the gasoline blending recipe optimization are applied as examples to prove the effective of the algorithm proposed in this paper.Experimental data approves that the adaptive genetic algorithm based on crossover library and parallel mutation can well overcome the problems of slow convergent speed, poor convergent precision and local convergence, and effectively improve the overall performance of genetic algorithm. |