Research On Super-network Modeling And Evolutionary Optimization Method For System-of-systems Artchitecture

With the development of informatization and intellectualization of weapon systems,the combat mode of modern warfare has changed dramatically.The connectivity between weapon systems is more frequent,and the characteristics of systematic combat and networking become more and more obvious.Especially,the emergence of unmanned equipment and swarm warfare styles makes the trend of systematic combat and networking more prominent.Under the background of this era and operational requirements,the traditional consideration of the merits of the single system performance,the effectiveness of single equipment has been unable to meet the operational needs,which requires that we must stand at the height of joint operations,from the perspective of System-of-Systems(SoS)construction and development,to introducte network modeling method,SoS-level simulation,intelligent calculation.Based on these techniques,we can scientifically optimize the configuration of combat SoS,and seek opportunities for future warfare under joint operational conditions.The architecture reflects the configuration of components in the SoS and the interaction between components and components and the external environment.It carries the physical entity,information structure and SoS capability,and is the core framework of SoS.The functional process of the architecture runs through the life cycle of SoS planning and design,demand demonstration,prototype development,application testing and actual application.Therefore,the thesis sets the research object as the combat SoS architecture,and realizes the optimal configuration of the combat SoS through the appropriate formalization expression and optimization of the combat system architecture.Aiming at the optimization problem of the combat SoS architecture,this thesis proposes the modeling and optimization framework.First of all,based on the networking characteristics of SoS and SoS architecture,the SoS architecture is modeled on the network,and the objective recognition of SoS architecture is unified,which forms the basis for the subsequent optimization.Secondly,the SoS effectiveness simulation data analysis method is used to acquire the architecture knowledge.Based on the hypernetwork model and the experience knowledge of the architecture,the many-objective optimization problem(MaOP),the target number is greater than or equal to three,of architecture optimization is constructed,and a preference-inspired co-evolutionary algorithm(PICEA)based on preference information is proposed to solve the architecture optimization problem to provide valuable auxiliary decision information for the operational decision makers.The main work and innovative contents of the paper are summarized as follows:(1)An architecture modeling and optimization framework for architecture design is proposed.The thesis analyzes the shortcomgings of existing architecture design methods in architecture modeling and optimization,puts forward the "network modeling and simulation data mining+architecture scheme optimization" new architecture modeling and optimization framework,and solves the SoS architecture optimization from three aspects of architecture super network modeling and simulation data mining based on computational experiments,and architecture optimization based on many-objective evolutionary algorithm.Among them,the architecture supernetwork model is the foundation of the architecture optimization,and provides the formal problem description.Simulation data mining based on computational experiments is a strong support for architecture optimization by providing architecture prior knowledge in multi-mission simulation environment.The many-objective evolutionary algorithm is the core of architecture optimization through providing the method flow and optimization algorithms.(2)The SoS architecture hypernetwork model is designed and constructed.Although the traditional view-based architecture modeling method can reflect the composition of architecture elements and their corresponding relations,it cannot reflect the dynamic evolution characteristics of the networked architecture.Based on the mechanism of SoS capability generation,the thesis constructs operational activities sub-networks,functional equipment sub-networks and organizational structure sub-networks.Meanwhile,accordingto the logical interaction between operational activities,functional equipment and C2 units,we conduct the super edge connection between the subnetworks.Then the super-network model of the combat SoS architecture with multiple structural levels,a large number of heterogeneous nodes and multiple attribute connections is generated.On the basis of constructing the super-network model,the paper proposes an architecture super-network model generation algorithm,and provides a method for model auto-generation,which further supports the coding and optimization of the architecture solutions.(3)The SoS performance simulation data analysis research based on the calculation experiment is realized.The SoS capacity generation process with many impact factors directly leads to the exponential increase of the architecture solution space,and the SoS model is also increasingly complex,which makes the architecture simulation optimization method adopting the solution traversal cannot be applied to practical applications.In this paper,the space of the simulation scheme is greatly reduced by the calculation of the near-orthogonal Latin hypercube(NOLH),and the effectiveness simulation data can be obtained by Monte Carlo simulation.The thesis then uses the regression model of the decision tree to perform regression analysis on the simulation data,obtains valuable auxiliary decision information,and supports the next step of the architecture optimization as the prior knowledge of the architecture.(4)The process of the optimization method of the SoS architecture and the many-objective evolutionary algorithm are proposed.Based on the super-network model and the prior knowledge of the architecture,the thesis comprehensively considers the multiple performance objectives such as the total performance level of SoS,the mission completion rate,the SoS completion time,the total cost,and the organizational structure load to build the Optimization Problem of Combat System-of-Systems Architecture(CSoSAOP).At the same time,in order to better solve such optimization problems,the thesis adopts Preference-Inspired Co-evolutionary Algorithm with goal vectors(PICEA-g)and designs a local principal component analysis operator(Local Principal Component Analysis,Local PCA)to improve the search ability of the algorithm on a high-dimensional space plane.Finally,the paper demonstrates the advanced nature of the algorithm on the benchmark set and the effectiveness of the algorithm on the battle scenario through two sets of case demonstration experiments.