Research On Damage Pattern Recognition Technology Of Space Debris Impact On Typical Spacecraft

As long as the development of aerospace industry,more and more space debris is left on the earth orbit.Especially the millimeter fragment called dangerous debris,which hard to elude by maneuvering in changing trajectory,and hard to defense by protective structure.Dangerous debris is the main threaten to spacecraft,but in particular to long-period and highly reliable spacecraft,such as space station.The on-board monitoring technique is presented in order to solve such problem,main functions include: whether be impacted so called ‘judge',locate the impact coordinate so called ‘locate',and speculate the damage pattern and level so called ‘damage'.The on-board monitoring based on acoustic emission is one of the most popular concerns by academic community and engineering department.So far,‘judge' and ‘locate' of on-board monitoring technique is tended to be mature,they can be applied for most flat and regular cylinder surface of spacecraft wall panel structure.However,research for ‘damage' is relative lack.Existing methods is restricted by experimental environment,size of ground simulation experimental element,structure and materials and so on.However,there are still several problems such as poor identification and low calculating speed.It's the key problem to be solved in on-board monitoring technical engineering process.In the light of of the problems above,to adopt the way by combination of ground simulation and emulation,based on acoustic emission technique,mainly study the research effort such as source location on comfortable structure,characteristic parameter design for representing damage,damage pattern recognition and damage level estimation,engineering applied technology,and so on.A damage pattern recognition module,which can recognize damage pattern of typical spacecraft wall panel impacted by space debris,and then estimate its damage level,is built.The main research is as follows:Based on obtaining the tendency of signal characteristics parameters,multiscale and time-window wavelet filtering method,which can remove the reflection signal,are observed.The relationship between characteristics parameter of space debris hypervelocity impact acoustic emission,such as time-frequency,energy,wave mode,and condition such as propagation distance,impact loading(impact velocity and projectile diameter)is obtained.On this basis,by means of build the relationship between time-frequency and wave mode characteristics,aim at the different scale,a propagation distance adaptive wavelet filtering method of multi-frequency-mode is designed.With this method,the damage pattern recognition technique can break through the reflection information aliasing limitation,which bring by size effect of ground simulation experimental element,for analyzing the damage pattern and damage level quantitatively,lays a technological foundation for common and effective characteristics parameters,which can be independent of specimen size.A zone-coordinate location algorithm module is built,which contains double automatics arrival time algorithm and virtual neighbors location algorithm.Processing isometric partitioning on acoustic emission signal sample sequence,the arrival signal piece can be obtain by using piece arrival threshold and piece adjust threshold,and then the double automatics arrival time algorithm is designed by searching arrival time in this piece.Based on the distance location method,further on partitioning of location training sample set by transducers group,the calculated amount could be decreased.Then a fusion switch method is used to solve the boundary problem on location net,and design a distance weighing based virtual neighbors location algorithm,by creating virtual sample with existing ones.Low velocity impact experimental result on a Shape of manned capsule shows that,this location algorithm have high calculating speed and accuracy.Under the certain training intensity,larger the spacecraft structure,more sensor network node,it will have higher mean positioning accuracy.The arrival time and positioning coordinate of this location module can be essential data for extract time and distance relevant characteristic parameter.Base on the Kruskal-Wallis test and optimize characteristics parameter combination,a damage pattern recognition module which contains relevance vector machine(RVM)based damage pattern recognition,and neural network based damage degree estimation.Base on location algorithm module,use time-frequency parameter,wavelet norm parameter and propagation distance to be the characteristics parameter,analyze the otherness between characteristics parameter and damage pattern and damage degree by Kruskal-Wallis test,and then make them to be the input of damage pattern recognition module.Result of ground simulation verification experiment shows that,module recognition rate for aluminum plate and honeycomb core sandwich is 90% and 96%,respectively,and estimation error of crater damage and hole damage can be lower than millimeter level.But more importantly,damage pattern recognition based on RVM method can break though the limitation that lack of ground simulation experimental sample.This module build on artificial intelligence(AI)method,solved the limitation problem to damage pattern recognition which contains experimental environment,test piece size,structure and materials and small sample,effectively.It's the key point to implement the hypervelocity impact damage pattern recognition engineering approaches for typical spacecraft.Location error-tolerant algorithm and modified model of damage pattern recognition are proposed.Based on eliminating the failure node by real-time failure diagnosis,we present the location error-tolerant algorithm by using residual sensor network.The verification result of low velocity impact on a manned capsule appearance module shows that,the error-tolerant ability of virtual neighbor algorithm is better than area location algorithm.Numerical simulation method,data fusion method and double threshold judgment for feedback standard error level,are used to amending the input parameter,decision-making process and damage degree result,respectively.Build a damage pattern recognition modified model,which advance the damage pattern recognition precision,and avoid the dangerous failure situation to the full extent.Location error-tolerant is the response strategy for sensor module on-board malfunction,modification to damage pattern is the key point to improve the sensor module ability,these two method are both important in damage pattern recognition in engineering process.In conclusion,we research several problems in engineering process of hypervelocity impact damage pattern recognition.On the basis of damage pattern relevant parameter and location module,a damage pattern recognition module and its engineering technique are built for typical spacecraft structure,which work fast and have high precision,and then verify the effectiveness with ground simulation.The achievement can solve the typical spacecraft damage pattern recognition problem,it break though the limitation of experimental environment and structure,and can adapt to complex structure and composite material,have a certain reference value to damage pattern recognition engineering technological application.This module can provide the necessary support to developing on-board monitoring which contains judge,location and damage modules.