Research On Dual Domain Robust Optimization Method For The Submarine Operational Performances

The submarine demonstration is at the beginning of its whole life cycle. In particular, the demonstration of submarine operational performances is the key node in the whole demonstration process, which as a bridge connects military requirements and conceptual alternatives. Hence, the resulting operational performances should both satisfy top military requirements and be feasible in submarine engineering. But the current demonstration methodology has some limitations that lead to two questions. One question is that whether or not the resulting operational performances can satisfy top military requirements optimally. The other is that that whether or not the resulting operational performances would be feasible in practice. The current demonstration methodology can not answer the questions satisfactorily. Therefore, at the latter stage it is possible that some decisions made before would be found false. But at that time adjusting the decisions would be very costly. Hence, it is a very important problem to identify the submarine operational performances based on top military requirements reasonably, and at the same time to keep the resulting operational performances stable.In order to assure that the resulting operational performances have considerable stability, this dissertation define the concept of dual domain robust optimality, and bring forward a new demonstration pattern for submarine operational performances to avoid some flaws in the traditional pattern. The main contents and achievement of this dissertation are listed as follows:(1)Aiming at the issue of submarine operational performances demonstration, a methodology of Multi-Expert based requirement-Traceable dual-domain Robust Optimization (METRO) is put forward. METRO emphasizes the traction from top military requirements. That is, the identification and the validation of operational performances are both based on top military requirements. Firstly, it constructs the traceable mapping between top military requirements and operational performances, and the cognitive robustness to this mapping is assured through the information fusion of multi experts' preferences. Then, It constructs the mapping from operational performances to conceptual alternatives, and alternative's robustness is assured through the uncertainty analysis. At last, an optimization process is used to integrate the above elements as a whole in harmony. Summing up, this methodology can resolve the feasibility, optimality, robustness and traceability of the resulting operational performances in a systematic manner.(2)Aiming at the issue of constructing the traceable mapping between top military requirements and operational performances, a method for the mapping of military requirements based on QFD/ANP and fuzzy integral is proposed. Firstly, based on QFD driven by customer needs and ANP, an integrated operational requirement analysis method is proposed, which can consider the complex relations between military requirements, operational tasks, operational performances, similar equipments, and conceptual alternatives simultaneously, and can handle the information in an uniform and structural manner. Hence, the nicety and consistency of the experts'judgement is improved. Thus, the mapping from top military requirements to operational performances is built. Then, the concept of fuzzy measures is used to describe the relations between the operational performances, and a new algorithm for the calculating of fuzzy measures is developed by the integration of ANP and Influence Matrix, which is suit for the situations without historical data, and is easily understood and executed. With this algorithm, the difficulties in applying fuzzy measures in MCDM are greatly decreased. At last, the military utility model of submarine conceptual alternatives based on Choquet fuzzy integral is established.(3)In order to assure the judgement about the mapping between between top military requirements and operational performances to be robust, a method of information fusion based on multi-expert multi-format preference and consensus reaching is proposed. Firstly, the methods to uniform multi-format preferences are studied. Especially the method to uniform multi- granularity and multi-semantic linguistic matrices is studied, which proves that there is no information loss and the linguistic comparison matrix will keep its properties after unification. Thus, by these unification methods, an expert would express his real preferences using the familiar format, which would improve the robustness of decisions. Then, a new method to deal with the interactions between the experts'preferences is proposed, which is based on the resemblance degree between the experts'knowledge and between the experts' knowledge comparison matrices, to calculate the 2-additive fuzzy measures to represent the importance of experts. Choquet integral is used as the aggregation operators to obtain group's preference. By this method, experts'power in decision making problem could be identified reasonably, which would be beneficial to obtain sounder group preference and to improve the robustness of decisions. At last, based on the evaluation of the aggregation and the measurement of group consensus degree, group consensus is reached through feedback information.(4)In order to quantify the influence of uncertainty from submarine sub-systmems, a method for the evaluation of robustness index based on Gaussian Process (GP) surrogate model is proposed. Firstly, the concept of robustness index of submarine conceptual alternatives is defined, and the robustness index of a single alternative can be calculated by Monte Carlo approach. But, calculating the robustness index of many alternatives will lead to a serious issue of calculating efficiency. To resolve this problem, a method for the evaluation of robustness index based on Gaussian Process (GP) surrogate model is proposed, which construct a response surface in the design space to fit the values of robustness index. Through this method, the robustness index value of arbitrary alternative can be evaluated quickly. In the process of GP surrogate modeling, the difficulties arise in the selection of core functions and the identification of hyper-parameters. Hence, this dissertation propose to use mixed core functions, and a new hyper-parameters optimization algorithm based deterministic annealing is developed, which can effectively improve optimization efficiency and results'precision.(5) Based on the work above, it is proposed to adopt the dual response surface based optimization approach to find the resulting operational performances that is optimal. A multi-objective optimization algorithm based on marginal distribution estimation is applied in the optimization of submarine operational performances. At the end of this dissertation, a example of a multi-task attack submarine is illustrated, which is used to validate the proposed methodology.