关键词: 航空复杂装备交付, 多团队分阶段共识, 数据驱动决策方法, 灰狼算法, 双层规划

Abstract: Delivery system of aviation complex equipment is an important means to realize corporate profits and ensure the sustainable combat capability of the military, but the delivery system for aviation complex equipment has not been established, which restricts the development of complex equipment development. To establish the delivery system, the primary task is ensuring the objectives, like the delivery cycle, customer satisfaction, and claims. The complex equipment's delivery system is composed of many important components, like the delivery system. Among the components, the objective of the delivery system is one of the most important parts. In previous works, the leader always accumulates all work, so that the whole process is a push-type construction. A push-type results in low efficiency. To avoid this shortcoming, we design a pull-type construction, which could enhance efficiency. In the pull-type construction, the objective is the primary thing, hence, we should define the objective of delivery first. The delivery concerns several groups, including the delivery and customer teams. The objective of delivery could only be obtained after a full discussion in multiple groups. However, it is difficult to realize the consensus of two different groups. On one hand, two diverse groups play different roles. Generally, a customer team is highly more important than a delivery team. On the other hand, the design of the objective is different from the selection of alternatives. The parameters of the designing objective are continuous, which is different from the selection of alternatives.
To address these issues, this study proposes the following innovative solutions: First, a dual-layer structural model is established to characterize team relationships. The client team is positioned in the first layer and the delivery team in the second layer, reflecting a client-first principle. Second, targeting the research gap in delivery parameter optimization, a data-driven group decision-making method is introduced. Considering the inefficiency of large-scale team negotiations, an intelligent evaluation method based on Recurrent Neural Networks (RNN) is designed. Compared with traditional Multi-Criteria Decision Analysis (MCDA) methods, this approach not only handles complex nonlinear relationships between attributes but also effectively captures cumulative effects in evaluations, significantly enhancing assessment efficiency. For model solving, an improved Grey Wolf Optimization (GWO) algorithm is proposed to address the nonlinear characteristics of the dual-layer programming model. By incorporating a Lévy flight mechanism to avoid local optima and integrating a gravitational search algorithm to enhance global search capabilities—particularly boosting local search performance—this method effectively resolves the inefficiency issues of traditional exact algorithms. To verify the good performances of the proposed method, a case study, which indicates the delivery of complex equipment, is used. According to the results, we could know that the predicting performance of recurrent neural networks is better than that of neural networks, which verifies the proposed conclusion. Meanwhile, the comparison between our proposed improved gray wolf algorithm and some heuristic algorithms shows the advantages of our proposed algorithms. The proposed mechanism in the gray wolf algorithm is useful.
Considering the features of fuzzy numbers, other kinds of predicting methods are encouraged to predict the experts' preferences. Also, as the delivery system is composed of kinds of different parts, it is worthwhile to study how to assign the objective value to each system.

Key words: aviation complex equipment delivery, consensus reaching process, data-driven decision making, gray wolf algorithm, bi-level programming