Experimental and numerical investigations are presented for a theory-guided machine learning (ML) model that combines the Hashin failure theory (HFT) and the classical lamination theory (CLT) to optimize and accelerate the design of composite laminates. A finite element simulation with the incorporation of the HFT and CLT were used to generate the training dataset. Instead of directly mapping the relationship between the ply angles of the laminate and its strength and stiffness, a multi-layer interconnected neural network (NN) system was built following the logical sequence of composite theories. With the forward prediction by the NN system and the inverse optimization by genetic algorithm (GA), a benchmark case of designing a composite tube subjected to the combined loads of bending and torsion was studied. The ML models successfully provided the optimal layup sequences and the required fiber modulus according to the preset design targets. Additionally, it shows that the machine learning models, with the guidance of composite theories, realize a faster optimization process and requires less training data than models with direct simple NNs. Such results imply the importance of domain knowledge in helping improve the ML applications in engineering problems.

Using two different methods, a previously developed elliptic blending model (the original SST k-ω-φ-α model) is modified for sensitization to streamline curvature. One method involves modifying the dissipation term in the turbulent dissipation equation, while the other constructs a new formulation for the turbulent kinetic energy production term based on an explicit algebraic stress model. The capabilities of the proposed models are evaluated by applying them to three flows with curved surfaces; namely, the two-dimensional (2D) infinite serpentine passage flow, the 2D U-turn duct flow, and the 2D periodic hill flow. The SST k-ω model with rotation and curvature correction (the SST k-ω-CC model) is also used for comparison. The computed results are compared with the relevant direct numerical simulation, experimental, and large eddy simulation data from the literature. It is found that the two proposed models significantly improve upon the original SST k-ω-φ-α model. Compared with the SST k-ω-CC model, the two proposed models produce better results in the 2D infinite serpentine passage flow and the 2D periodic hill flow. The proposed models are similarly competitive with the SST k-ω-CC model in the 2D U-turn duct flow.

No abstract

Ultra-high-performance concrete (UHPC) is expected to provide solutions for the development of lightweight, high-strength, and rapid construction of concrete bridges due to its excellent material properties. In order to study the influence of steel fiber on the bending performance of non-reinforced UHPC (NR-UHPC) slabs, the bending failure test of 8 NR-UHPC slabs was completed with the steel fiber content as a parameter, and the failure mode, load-deflection, crack widths and load-strain curves of the test slabs were analyzed. The test results show that the destruction process of the NR-UHPC slabs can be divided into three stages: elastic stage, cracking stage and failure stage. The load-deflection curve and load-strain curve of the specimen with steel fiber content of 0.5%-1% change linearly during the elastic phase. In the elastic phase, the load-deflection curve and the load-strain curve of the test piece with a steel fiber content of 2%-3% were not smooth after the linearity begins; the strain of the UHPC slab was almost the same within 0-10 kN. However, as the load increased, the larger the amount of steel fiber was, the smaller the strain was, that is, the UHPC slab had good crack resistance. According to the regression analysis of test results at home and abroad, recommended design formulas for the bending bearing capacity of NR-UHPC slab were put forward, which tally well with the test results.