Study On The Manufacture And Performance Optimization Of Reinforced LVL

In the paper, three different kinds of poplar (Populus ussuriensis Kom.) laminated veneer lmber (LVL) were manufactured, including the nonreinforced LVL, the LVL reinforced with monolayer fiberglass mesh and the LVL reinforced with multilayer fiberglass mesh. The study on the nondestructive testing, the optimization of production technology and the reinforcement design was implemented.The dynamic Young's moduli of the three kinds of LVL were obtained by using the nondestructive testing methods, including the Ep, dynamic Young's modulus by longitudinal vibration method, the E1, dynamic Young's modulus by out-plane flexural vibration method, and the E2, dynamic Young's modulus by in-plane flexural vibration method. The values of static modulus of elasticity (MOE) and static bending strength (or modulus of rupture, MOR) were also obtained by using the static bending test. The linear correlativitys were investigated between the dynamic Young's moduli and the MOE or MOR for the three kinds of LVL.For the vertical loading test specimens, the linear correlations between Ep, E1 and MOE or MOR were good for the nonreinfored LVL; the linear correlations between Ep, E1, E2 and MOE or MOR were also good for the LVL reinforced with monolayer fiberglass mesh; the linear correlations between Ep, E1, E2 and MOE were good for the LVL reinforced with multilayer fiberglass mesh, but the linear correlations between Ep, E1, E2 and MOR were poor.For the parallel loading test specimens, the linear correlations between Ep, E1, E2 and MOE or MOR were good for the nonreinfored LVL and the LVL reinforced with monolayer fiberglass mesh; the linear correlations between Ep, E2 and MOE or MOR were good for the LVL reinforced with multilayer fiberglass mesh.In the paper, the optimum design for production technology of the nonreinfored LVL was implemented by using the models of quadratic polynomial regression equation and BP (Back Propagation) neural network. The prediction models were established between the production technology parameters and the MOE or MOR. The prediction value of the press temperature was the same for the two models. But the prediction values of the press duration and the adhesive spread were different for the two models. The variation range of prediction values for MOE and MOR using the BP neural network model was smaller than that of prediction values using the quadratic polynomial regression equation model.The reinforcement design for the LVL with fiberglass mesh was also implemented in the paper. The reinforcing effect of lay angle and adding position of the fiberglass mesh was studied. The reinforcing effect was very good for the vertical loading test specimens reinforced with monolayer fiberglass mesh. When the lay angle was 30°, the reinforcing effect for MOE and MOR was better. The reinforcing effect was also very good for the vertical loading test specimens reinforced with multilayer fiberglass mesh, with the range from 20.38% to 41.70% for MOE and 29.46% to 40.10% for MOR. But the MOE and MOR were weakened for the parallel loading test specimens reinforced with fiberglass mesh. So the LVL reinforced with fiberglass mesh should be used in the occasion where the loading direction was perpendicular to the glue-line direction, not be used in the occasion where the loading direction was parallel to the glue-line direction.The test scheme was designed with the orthogonal test method for the LVL reinforced with multilayer fiberglass mesh. The crevices between any two wood veneers were designed as the experimental factors, and the experimental levels were either adding or no adding the fiberglass mesh in the crevice. The direct analysis and the variance analysis were used to analyze the orthogonal test results. The prediction model between the experimental factors of adding position and the mechanical properties of the reinforced poplar LVL was constructed with the BP neural network. This prediction model was used to find the best adding position of multilayer fiberglass mesh in all combination of factors and levels. The prediction results of the variance analysis and BP neural network model were compared with each other. The prediction results of the two methods about the best adding position were basically consistent. The best adding position was that the fiberglass mesh should be added to the middle crevices of LVL.