| Summary: | 碩士 === 國立臺灣科技大學 === 機械工程系 === 96 === Spinal pedicle screw has two important design objectives to be considered: bending strength and pullout strength; these two objectives may conflict with each other. The present study reported how to find the optimum solution of the multiobjective problem in the spinal pedicle screw with 7.5mm outer diameter. Finite element analysis and Taguchi L25 Orthogonal Array were used to develop a data set for learning process of artificial neural network (ANN). The neural network then gave surrogate models which were used as the cost functions in the optimization design. Then, the trade-off solutions known as Pareto optima were explored by a weighted-sum aggregating approach and the genetic algorithm. The ANN models could accurately predict the results of finite element analyses. The optimum design of the spinal pedicle screw was 0 mm for the initial position; 4.465 – 4.903 mm for the inner diameter; 0.4 mm for the proximal root radius; 2.884 to 2.961 mm for the pitch; 50 for the proximal half angle; and 0.1 mm for the thread width. In conclusion, the ANN could reliably predict the results of finite element analyses and the multiobjective optimization solution with genetic algorithm could be used to obtain the optimum design of pedicle screw. This method could reduce the time, cost, and labor and contribute to the manufacturers and surgeons in developing and selecting an ideal design for the patients.
|
|---|
