A method of optimizing the drum-shape design and assembly performance of a compressed rubber torsional vibration absorber(TVA) was presented. First

the criteria for assessing the assembly performance and the approaches to experiment the criteria were introduced. Then a calculation of the criteria using the finite element method(FEM) was developed. The results of the calculation were compared with the experimental ones to validate the developed calculation. Since the influence of the drum-shape of a TVA on its assembly performance has more than one factor

the orthogonal experiment is used to analyze the influence of each factor on the assembly performance. Finally

the method of optimizing the drum shape design and maximizing the assemble function of a TVA is proposed based on the BP neural network along with genetic algorithms. The optimized results are compared with the experimental ones

and the relative errors are within 10%

which validates the proposed method.