Improvement Of BP Algorithm Based On Fusion Algorithm And Its Application In The Supportability Assessment

Integrated equipment support (Integrated Logistics Support) job is to ensure that the newly developed equipment can be easy to guarantee effective protection when it is brought into use, which is the main ways of enabling rapid formation of combat and reducing the life cycle cost. Mainly to resolve the equipment life cycle process and security-related issues, providing support resources related to the equipment and the establishment of the matching protection systems to meet readiness requirements of peacetime and wartime. Come out from the equipment support,equipment support assessment become the focus of equipment support projects.However, artificial neural network has a unique association, memory and learning functions, parallel processing, fault tolerance and arbitrary function mapping capabilities, the use of powerful learning skill of neural network system can solve the problem.Therefore, the artificial neural network theory is applied to equipment support for evaluation can be achieved based on the value of experience and knowledge in many neural networks, such as BP (BackPropagation) network, the most widely used network,which has been used in BP algorithm become the most widely used neural network learning algorithm, the vast majority of neural network models are based on the BP algorithm or a variation of it, but people has revealed the existence of limitations of such algorithm in the process of using it, such as the convergence rate is too slow, easy to fall into local minimum values, and trend of forgeting the old sample. These limitations seriously affected the performance of BP algorithm.In this paper, according to the shortcomings of BP algorithm,improve the BP algorithm from the two areas of initial weights optimization and internal BP improvements.In the internal BP, study the momentum term and adaptive learning rate of BP algorithm, improving the BP algorithm converges slowly and shocks and other defects.In weight optimization,improved from integration of evolutionary algorithm and BP algorithm, we propose a improved BP neural network algorithm model based on fusion algorithm.The algorithm solves equipment supportability assessment questions in the case whether exits expert guidance.If it has expert guidance,we will use AHP algorithm combined with expert guidance in determining the initial weights of BP network,without expert guidance,we use ant colony algorithm and genetic algorithm optimization the initial weights.In this case,Imitate the ant colony algorithm solving TSP question,we propose network ant colony algorithm to solve the question of initial weights of BP network.And in accordance with MMAS (max min ant) system improve the ant colony algorithm.And analyzed the genetic algorithm and AHP hierarchical algorithm for the initial weights of BP.Finally,validate the fusion algorithm model by equipment supportability engineering data.