Model And Data Hybrid-Driven Approaches For Weapon Equipment Assessment

As one of the basic tasks for research on weapon system-of-systems,the assessment of weapon equipment can provide quantitative indices for the subsequent architecture design and optimization.For a long time,model-driven assessment methods based on expert knowledge in related fields have been widely used as the mainstream method for weapon equipment assessment.However,due to the high complexity and uncertainty of weapon equipment,it is difficult for researchers to establish an accurate and universal assessment model for the assessment of weapon equipment.With the rapid development of information technology and data science,data-driven assessment methods for weapon equipment have emerged.The whole processes of weapon equipment assessment,such as model establishment,parameter setting,and inference,depend on objective data in the data-driven assessment methods,which enhances the accuracy of the assessment results.The data-driven methods of weapon equipment assessment also provides a new solution for effectively solving the above-mentioned problems.Based on the analysis of the characteristics of weaponry equipment assessment,it is believed that there is a need to effectively integrate the expert experience,historical information,objective data,and other information to carry out weapons equipment assessment.Therefore,the model and data hybrid-driven approaches are more applicable to weapon equipment assessment.In this dissertation,multiple information fusion in different scenarios of weapon equipment assessment is regarded as the core problem.The basic idea of describing and processing uncertain and incomplete information in weapon equipment assessment with belief rule bases is proposed.The applicability of belief rule bases in weapon equipment assessment is also analyzed.Then,the framework of model and data hybrid-driven approaches is provided on the basis of belief rule bases.Furthermore,to address the problem of high complexity in the learning process of belief rule bases,a series of model and data hybrid-driven approaches for weapon equipment assessment are proposed according to the actual requirement in a variety of scenarios.The major contribution and innovations of this dissertation can be concluded as follows:(1)A technical framework for the model and data hybrid-driven assessment of weapon equipment on the basis of belief rule bases.Model-driven is a widely used method for the analysis and demonstration of weapon equipment based on expert knowledge.Data-driven is a new and effective method to process objective data of weapon equipment.Based on the analysis of characteristics and related concepts of weapon equipment and weapon equipment assessment,relevant information and data including expert knowledge,historical information,and objective data are defined as generalized information for weapon equipment assessment in this dissertation.It is also believed that the model and data hybrid-driven method is effective to deal with the generalized information in the weapon equipment assessment.On this basis,this dissertation focuses on the multi-information processing and fusion in weapon equipment assessment.Based on the analysis of the mechanism of the belief rule bases inference and parameter learning,this dissertation is committed to the description and processing of the expert knowledge,historical information,and objective data in the weapon equipment assessment.An overall framework is also proposed for the model and data hybrid-driven assessment of weapon equipment,which is developed to effectively use equipment data and expert experience in the assessment of weapon equipment.(2)An approach for weapon equipment assessment with complete informationThis dissertation strikes to propose a model and data hybrid-driven approach to improve the accuracy of the weapon equipment assessment,whose core is to adjust the model settings of weapon equipment assessment through the inference and parameter learning process of belief rule bases.However,the complexity of traditional approaches for the construction and parameter learning of belief rule bases is relatively high,which seriously affects the application of this method in actual problems of weapon equipment assessment.To address this problem,an approach for weapon equipment assessment with complete information is proposed in this dissertation.In this approach,the Causal Strength Logic is used in the construction and parameter learning process of belief rule bases.The parameters of Causal Strength Logic are first set and optimized before calculating to construct the belief rule base with Causal Strength Logic algorithms.With this regard,the complexity of the construction and parameter learning of belief rule base can be reduced without lessening the accuracy of weapon equipment assessment.A case study of fuel delivery equipment capacity satisfaction assessment is studied to verify the effectivity of the proposed approach.The approach for weapon equipment assessment with complete information are applicable to the assessment problems with sufficient information and are suitable for constructing complete structural belief rule bases.(3)An approach for weapon equipment assessment based on reduction of complete informationIn the approach for weapon equipment assessment with complete information,the belief rule bases for weapon equipment assessment are constructed under the conjunctive assumption.Namely,the antecedent parts of belief rules are the combination of possible referenced values of all the antecedent attributes.As a result,it always emerges a combinatorial explosion problem of antecedent attributes when applying the belief rule bases under the traditional conjunctive assumption in multi-attribute assessment problems.To address this problem,an approach for weapon equipment assessment based on reduction of complete information is proposed.Compared with the approach for weapon equipment assessment with complete information,the belied rule bases are constructed under the disjunctive assumption.First,the disjunctive assumption of belief rule bases is defined in this dissertation.The relevant activation and aggregation algorithm of belief rules in belief rule bases under the disjunctive assumption are also provided.Second,a bi-level algorithm of belief rule base optimization is proposed by integrating the rule numbers and accuracy in a united objective function with Akaike's information criterion.The structure and parameter settings can be determined through the bi-level algorithm of belief rule base optimization.At last,a case study of fuel delivery equipment capacity satisfaction assessment is studied to verify the effectivity of the proposed approach.In the approach for weapon equipment assessment based on reduction of complete information,the belief rule bases under the disjunctive assumption require less information for the assessment of weapon equipment.Therefore,this approach is applicable to the assessment of more general weapon equipment than that of the approach for weapon equipment assessment with complete information.(4)An approach for weapon equipment assessment with incomplete informationIn most actual problems of weapon equipment assessment,it is difficult to acquire all the required information due to the high uncertainty of the combat filed environment and weapons development.It is of great significance to propose an approach for weapon equipment assessment with incomplete information.To address this problem,an approach for weapon equipment assessment with incomplete information is proposed.Based on the idea of self organizing map,the approach consists of three phases.First,subsets of belief rule bases are constructed for the weapon equipment assessment with local complete information.Second,optimize these subsets,and calculate the weights of subsets belief rule bases.Third,all the subset are aggregated to acquire the final belief rule base for weapon equipment assessment with incomplete information.A case study of local air defense capability assessment is conducted to illustrate the proposed approach,and the feasibility of the approach for weapon equipment assessment with incomplete information is also verified in this case study.