| In practical problems in modern engineering,information of dynamic loads is often used as a prerequisite for structural design and optimization,and is widely used in the automotive field,construction industry,and modern construction.Therefore,the study of dynamic load identification methods has significant value for engineering problems.In addition,due to the defects of the manufacturing process and the discreteness of the material itself,there are unavoidable random errors in the characteristics of the structures.Therefore,how to realize the load reversal under the non-deterministic system is more practical and challenging for modern engineering.At present,some research achievements have been made in the field of dynamic load identification of uncertain structures,but most of them are applicable to situations of low uncertainty level,or will cause low computational efficiency.When the variation of uncertain parameters is large,there will inevitably be deviation in the identification of dynamic loads by such methods.In addition,for structures with uncertain parameters with different probability distributions,the study of dynamic load identification methods is also very lacking.Therefore,it is necessary to develop a direction of dynamic load identification of uncertain structures on how to accurately identify dynamic loads with parameters of high uncertain level and different probability models.It is also an important development direction to develop a fast and efficient dynamic load identification technology in the future.Aiming at this direction,this article combines manifold analysis,regularization methods,sampling techniques,finite element methods,probabilistic modeling techniques and correlation modeling methods to study the dynamic load identification technology with uncertain parameters under different probability models.The effective inversion of the dynamic load of uncertain structure is realized under a large degree of uncertainty.Specifically,the main content of this article is as follows:(1)Based on the probability measure,a method for identifying dynamic loads of uncertain structures based on manifold analysis is proposed.The uniform sampling method is used to obtain the uncertain structural material parameter samples,and the regularization method is used to obtain the structural dynamic loads under each parameter sample.All the features of dynamic loads are mapped to the low-dimensional shape space by manifold analysis method,and the dynamic loads are then reconstructed from the features of the shape space to achieve dimensional reduction.On this basis,the relationship between the uncertain parameters and the load characteristics in the low-dimensional shape space is constructed,so that the dynamic load can be identified directly by the uncertain parameters.Using this method,there’s no need to solve the inverse problem by numerical model again,and the emergence of ill-posed problems is also avoided and the data processing efficiency is improved.In the examples of the truss and car bonnet,the dynamic load inverse is realized based on manifold analysis method under the probability measure,and the probability density functions of the dynamic loads are obtained as the evaluation index.(2)The compound antithesis method of structural dynamic load and structural parameters based on ellipsoid model is studied.When there is less information about the structural uncertain parameters,that is,when only the boundaries and correlations of the uncertain parameters are obtained,the ellipsoid model is used for description.Considering the local damage of the structure,the composite reverse calculation of structural dynamic load and structural parameters is carried out.According to the kinematics equation,the internal force caused by structural defects can be transferred to the right end of the equation,and then inverse model of the dynamic load can be established according to the transferred kinematics equation.After solving the dynamic load of the structure and the internal force caused by structural defects,the inverse of the structural parameters is realized by substituting the internal force of the structure into the kinematics equation,so as to realize the composite inverse of the dynamic load and structural parameters.By mapping the load characteristics to the low-dimensional manifold space,the relationship between the uncertain parameters in the ellipsoid domain and the load characteristics is established,and the effective inversion of dynamic loads and internal forces is realized.This method is not limited by the amount of information of the parameters,and can effectively realize the composite reverse calculation of dynamic loads and structural parameters.(3)Developed a structural dynamic load identification method based on highdimensional probability-ellipsoid mixed uncertain parameters.A dimensionality reduction stratified sampling method is proposed.By dividing the probability and the ellipsoid parameters into inner and outer layers,the load identification problem under high-dimensional uncertain parameters is transformed into two types of lowdimensional uncertain problems,that is,inversing the dynamic loads with the probability parameters of the inner layer and establishing the relationships between the ellipsoid parameters of the outer layer and the dynamic load cumulative distribution manifolds.The former transforms the high-dimensional mixed uncertainty problem into a low-dimensional one under a single probability distribution to solve,and obtains the dynamic load cumulative distribution curve under each probability parameter.On this basis,the latter establishes the manifold relationships between the ellipsoid parameters and the cumulative distribution curves of dynamic loads,thereby indirectly establishing the relationship between the dynamic load and all uncertain parameters.This method avoids the problem that a large number of samples are required for the high-dimensional uncertain inversion process,and realizes the effective identification of structural dynamic loads under high-dimensional mixed uncertain parameters. |
