Research On Composite Inversion Of Dynamic Loads And Structural Parameters Based On Sub-structure Analysis

Unknown input and unknown structural parameters often exist together. Inpractical engineering problem, the dynamic load of structure is often difficult tomeasure directly because of the economic or technological limits. And the structuralcharacteristic parameters are also difficult to obtain for some complex structures.Therefore, dynamic load and characteristic parameters identification is an importantproblem for modern engineering. For solving this problem, a composite inversionprocedure is proposed based on sub-structure analysis. The dynamic loads andparameters of the model are identified only by limited output observations. Maincontents of the dissertation are presented as follows:(1)A composite inversion algorithm based on interface forces was studied. Firstly,the structure is decomposed into several sub-structures according to the structurecharacteristic and the finite element model of the sub-structure with unknown input isbuilt. Its connecting relationship with others is replaced by the interface forces. Theforward analysis model of the dynamic load identification is established by Green’sfunction. Based on regularization method, the time history of dynamic load can bestably reconstructed. Then, the finite element model of the global structure is built onthe basis of identified dynamic loads, and the equation of the structure parameters isbuilt through nonlinear least squares method. The equation is solved by usingregularization method in the iterative process, which make the identification resulttends to be stable. The numerical example indicates that the composite inversion iseffective and stable for the loads and structural parameters identification with noisyresponse.(2)A composite inversion algorithm based on interface forced motion and modalanalysis was studied. The connecting relationship of the sub-structure with others wasreplaced by the interface displacement and velocity. The motion equation of thesub-structure was deduced and it was projected to the modal coordinates based on themode decomposition. Modal force of every decoupling equation is identified in modaldomain. The assembly of Modal force is projected to the physical coordinate systemto get external loads. This method can reduce the dimension of the kernel matrix andexpand the application scope of the algorithm.(3) In order to realize the reconstruction of the load stably and improvecalculation efficiency, a composite inversion algorithm based on shape function and BP Artificial Neural Network was studied. Firstly, the structure is decomposed intoseveral sub-structures. The motion equation of the sub-structure was built to identifyloads. Dynamic load is decomposed into small elements in the time domain. Everyshape function of the dynamic load was constructed by nonlinear least squares methodbased on base function. The responses of the shape functions was obtained throughfinite element method, and assembled to form the global shape matrix in the wholetime domain. The shape function forward model of load identification is established.This method can keep considerable accuracy for long sampling period, and increasethe smoothness of the load curve. Then, the identification modeling of structureparameters is built by BP artificial neural network. In this way, the nonlinear modeldoesn’t need to be approximated by using the first-order Taylor formula. Meanwhile,the iterative calculation process for the finite element model is not required andthe computational efficiency is improved.