Constrained Stochastic Imperfection Modal Method
Based on virtual interation forces, the structural topology constraint is considered, and the constrained stochastic imperfection modal method is proposed for single-layer reticulated shells with controllable computational cost and reasonable accuracy. Journal paper
Early-warning Collapse of Planar Steel Trusses
Through parametric analysis, all collapse modes and mechanisms of planar steel truss structures under arbitrary conditions are summarized, and an early-warning method for fire-induced collapse is proposed based on the evolution laws of key physical parameters (joint displacement and displacement velocities). Journal paper
DNN-driven Real-time Displacement Prediction
Based on the measured data of structural fire response and deep LSTM neural networks, the real physical model of the fire-stricken building is identified, and a real-time prediction method for hard-to-measure critical displacement of the building under fire is proposed. Journal paper Codes
Critical Element Identification of Frames
A quantitative approach is proposed for identification of critical element identification of frame structures against progressive collapse, based on deep graph neural networks and reinforcement learning. Click the figure to see the optimized demolition plan for a 4×5 frame. Accepted manuscript
Machine-specified Ground Structure Method
The topology of 2D truss structures can be self-generated using the graph embedding and the policy network. The self-generated topologies are further utilized to generate the machine-specified ground structure, which can be applied to solve topology optimization problems with singular optimal solutions. Click the figure to see the self-generation process of a truss based on a 6×3 node-set. Journal paper
Form-finding of Aluminum Alloy Reticulated Structures
A quantitative index to evaluate the constructional convenience of free-form aluminum alloy reticulated shells is proposed considering the details of the aluminum alloy gusset joint. A form-finding method of the free-form shells is proposed based on the index mentioned above. Journal paper
Prediction of Buckling Load of Imperfect Reticulated Shell using Machine Learning
The consistent imperfection method is modified according to the specification on the constructional errors. Based on the modified consistent imperfection method and supervised machine learning, a prediction method of the non-linear buckling load of imperfect spatial structures with low calculation cost is proposed. Journal paper
Shape Optimization for Aluminum Alloy Reticulated Shells
A shape optimization method of aluminum alloy shells considering joint rigidity is proposed by means of the genetic algorithm. The method can improve the non-linear buckling load of a spherical shell without significantly change its geometric appearance. Journal paper link