Improvement Of Genetic Algorithm And Using In Application And Research Of Scheduling Optimization

As an important part of the connecting management and production, production scheduling relates to the integral achievement of the enterprise resource planning (ERP) and production controlling. The mature and general production scheduling software can help enterprises improve productivity, reduce inventory costs and improve regulatory capacity. And the soul of the mature and general production scheduling software is the production scheduling technology. Plan Evaluation and Review Technique is an outstanding representative of the project which determines the level of profits. In our country, the scheduling optimization of network planning technique basically is achieved using the network planning technique of Operations Research or the mathematical programming which can provide the a good solution to the decision-makers. However, it is not be guaranteed the result is the best and a single solution cannot meet the various requirements of decision-makers in practice. This integer programming or linear programming method requires a lot of computation and the time complexity of computing is higher, searching efficiency is very low when the number of jobs in the project is too large or relation between jobs is complicated. In response to these problems, applying the existing intelligent algorithms to the network planning optimization has become one of the hot domestic for the scholars at home or abroad. The existing Intelligent algorithms essentially deal with discrete optimization searching problems, and they do not require the continuity of the problem space, have no gradient in formations, their robustness has been confirmed and the capacity of dealing with complex optimization problems has made significant achievements. So, they have more advantages than the others when to solve large-scale network planning in the multi-objective optimization problems. The genetic algorithm is one of the representatives of such intelligent algorithms.But, the simple genetic algorithm has two major flaws. Firstly, simple genetic algorithm can bring easily premature convergence; the other is the process for working out the optimal values using simple genetic algorithm is relatively slow. The simple genetic algorithm can effectively solve the single-extreme value optimization problem, but for multi-extreme optimization problem, it is often less than the global optimal convergence and in the local search capacity, the simple genetic algorithm is not very good. Using the simple genetic algorithm to slove practical problems cannot achieve the best global optimal solution but achieve the local optimal solution because of various reasons.For the above defects in the simple genetic algorithm, this paper presents an improved adaptive genetic algorithm. For the improved adaptive genetic algorithm, the crossover probability and mutation probability of the largest fitness chromosome is not zero, the crossover probability and mutation probability in the evolution are not unchanged, the mutation probability of the poor fitness and the diversity of population both have been improved, the easily premature convergence has been avoided and the global optimal values can be achieved.This paper applies the improved adaptive genetic algorithm to solve resource optimization, time optimization and cost optimization problem in network optimization planning scheduling optimization and Comprehensive optimization problem.(1)Resource optimization problem. This problem can divide into two categories:"fixed duration---resource balanced" and "limited resource---the shortest duration" problem. For the "fixed duration---resource balanced" problem, this paper puts forward the adaptive genetic algorithm design for single resource balanced optimization and multiple resource balanced optimization, then compares the results from the SGA with the results from the AGA. For the "limited resource---the shortest duration" problem, this paper refer to the method from the reference [1] and puts forward the adaptive genetic algorithm design and gives the authentication in comprehensive optimization with comparing the results from SGA with the results from AGA.(2)Time and cost optimization. According to the relationship between time and cost,the problem can be divided into two types:continuous time and cost optimization and discrete time and cost optimization. For the continuous time and cost optimization problem, this paper has put forward daptive genetic algorithm solution for two-goal that time and cost and comparing the results from reference [1] using the simple genetic algorithm with SGA this paper has put forward. The results show that the more efficiency to solve the time and cost optimization problem using adaptive genetic algorithm. For the discrete time and cost optimization problem, each work not only has its number but also has some construction methods that have consumption cost and resource. Its SGA genetic algorithms design is similar to the continuous time and cost optimization but gene encoding. This paper gives the authentication for discrete time and cost optimization in comprehensive optimization with comparing the results from SGA with the results from AGA.(3)Comprehensive optimization. The actual construction process of projects or tasks, we must consider not just unilateral optimization, often to time, cost and resource to integrate the three optimization objectives. Therefore, the actual project and its tasks requires integrated optimization for time, cost and of resource. Based on the reference [1], this paper has put forward the AGA design for comprehensive optimization. Firstly, combine the "limited resource---the shortest duration" problem with discrete time and cost optimization problem that as the first step of comprehensive optimization. To obtain the shortest duration and the start time of job information to resource balanced optimization module. Resource balanced optimization module uses the duration to calculate the latest start time of each work. Make the start time as the earliest time, then generating time between the earliest time and the latest time as the actual start time. Finally, output the results from the two optimization modules. Examples' results show that using the AGA to solve the comprehensive optimization problem has more efficient than the SGA that reference [1] has put forward, also shows that using the AGA this paper has put forward to solve the network planning scheduling optimization problem more feasibility and efficiency.Innovation of this paper is that improve the simple genetic algorithm, applying the improved genetic algorithm to the network planning scheduling optimization problem and comparing the results from reference [1] for the resource optimization, the cost optimization and the time optimization. The results show that using the SGA this paper has put forward to solve the resource optimization, the cost optimization and the time optimization problem is more efficiency than the SGA the reference has put forward and can converge to the optimal value, the converge rate is fast than SGA.