Dynamic Bayesian Network Modeling And Assessment Method Of Structural State Evolution For Reusable Spacecraft

With the completion of China’s long-term on-orbit space station and the gradual implementation of various space exploration and development tasks,large-scale and efficient access to space will become an inevitable demand in the future.The reuse of space structures,such as spaceships and launch vehicles,will significantly reduce the cost of space launch and return,and become the goal of the development of space powers.The American Dragon spacecraft has been reused,and China and Russia have carried out reusable technology verification in the new generation of spacecraft.Due to the particularity and high risk of the mechanical environment of space launch,there are many uncertainties in the repeated use of space structures,such as the mechanical environment and the evolution of structural characteristics.How to evaluate the state evolution and its uncertainty of reusable space structures is the key to the health management of reusable space structures in service.The digital twin technology measures the structural response in real time through the sensing system,and improves the accuracy and credibility of the structural state and its evolution prediction by improving the prediction model through introducing dynamic data,provides a new solution to the problem of accurate evaluation and decision-making of reusability of aerospace structures under strong uncertainty.In this paper,the main load-bearing structure of spaceship is taken as a typical research object,and the structural fatigue and bolt loosening under various dynamic loads experienced by spaceship during launch,landing and reentry are taken as two key factors affecting the reusable performance of the structure,and the digital twin modeling method of the main load-bearing structure is studied under the condition of uncertain model state and load input.Focus on structural life evolution modeling and prediction,response acquisition of key dangerous locations,and research on dynamic data-driven structural reusability prediction methods,as follows:(1)Aiming at the structural state and evolution uncertainty under fatigue damage,adopting a dynamic Bayesian network(DBN)to fuse multi-source uncertainty and establish a structural fatigue damage evolution prediction model.Furthermore,a variable structure DBN modeling method is proposed to solve the problem of structural multimode damage evolution modeling.The uncertainty of spacecraft structural life prediction is reduced by using observation data to drive DBN model updating and damage evolution prediction.Based on the stress-strength interference model and structural failure criterion,the analysis method of structural reliability is established,which is used as the criterion of structural reusability.By combining reliability prediction with DBN model,a dynamic assessment method for the life-cycle reliability of reusable structures is established.Particle filter is used for DBN reasoning.In order to solve the particle impoverishment problem,Metropolis Hasting sampling is introduced into the resampling algorithm,and new particles are added adaptively.At the same time,the model weighted fusion method is developed to improve the accuracy and stability of DBN reasoning.(2)The fatigue test data of standard specimens are used to verify the effectiveness of DBN in the prediction of structural fatigue crack growth.The DBN model for predicting the fatigue crack growth of the standard specimen is established,and the fatigue crack growth model state is dynamically modified by the crack growth observation data,so that the accurate prediction of the different growth processes of each specimen under constant amplitude loading and variable amplitude loading is realized.The reliability of the self-adaptive adjustment method for the model state update time point is verified.The research on the change of information entropy shows that the introduction of data can continuously reduce the uncertainty of the prediction results.Aiming at the main load-bearing structure of spacecraft,the crack propagation life prediction of the main load-bearing structure with only fatigue damage or with both fatigue damage and bolt loosening damage is studied.By defining the structural damage characteristic quantity and the stress amplification coefficient,and analyzing the probability of bolt loosening and its influence on crack propagation,the criterion of structural change of DBN network is established.The manufactured solutions of the above methods show that the traditional DBN can not effectively track the effect of accelerated crack growth after bolt loosening.Even under the condition of inaccurate bolt looseness criterion,the variable structure DBN can still accurately track the accelerated crack growth behavior through multiple data introductions,and has good scalability for life evolution prediction under multi-source damage.(3)Aiming at the uncertainty of input loads caused by the difficulty of directly measuring the key position responses of aerospace structures,the empirical mode decomposition of the measured signal and the modal superposition of the response to be reconstructed are used to establish the reconstruction method of the dynamic response of the target position using the limited measured data.Based on the adaptive noise complete set method,the modal aliasing problem of the traditional empirical mode decomposition in the response reconstruction of the main load-bearing structure is solved,and the response reconstruction accuracy is improved by combining with multi-point data fusion.The above method is used to reconstruct the acceleration and strain responses of the key positions of the main load-bearing structure of a spaceship under random vibration loads,and numerical and experimental verifications are carried out.The results show that compared with the traditional empirical mode decomposition,the reconstruction method based on complete ensemble empirical mode decomposition with adaptive noise improves the reconstruction accuracy by 80%;the more measurement points involved in the fusion,the higher the reconstruction accuracy is.(4)Aiming at the uncertainty of the model and input in the life prediction of aerospace structure,the digital twin framework of life and reliability prediction of the main load-bearing structure of spaceship is constructed by taking fatigue crack propagation as the main factor of structure life.Aiming at the problem that the structural damage evolution will change the dynamic characteristics of the structure and affect the accuracy of response reconstruction in repeated use,the reconstruction and crosscomparison of multi-point strain response are used to establish the identification method of structural modal changes caused by bolt loosening and crack propagation,and to drive the structural modal update,so as to reduce the response reconstruction error under the condition of structural dynamic evolution.The dynamic prediction of the crack growth life and the reliability evolution of typical structures is realized by taking the dynamic reconstructed key position responses as the input of the DBN prediction model and driving the DBN model update with the crack observation data.Compared with only considering the model uncertainty,the predictions considering both the load input and the model state uncertainty are in better agreement with the test results.The digital twin framework for the main load-bearing structure of spacecraft proposed in this paper solves the problem of large model state and input uncertainties in life prediction,and the effectiveness of the framework is verified by experiments.It can provide a theoretical method for the health management,reusability and mission decisionmaking of reusable spacecraft structures.