Ontological model-based optimal determination of geometric tolerances in an assembly using the hybridised neural network and Genetic algorithm

Abstract

Traditional Geometric tolerance allocation method assumes the parts of the assembly are purely rigid. Upon assembly and function, it is realised that the rigidity of the parts is non-ideal which leads to severe variations. Hence the performance of an assembly declines and is associated with the parts performance in the assembly. An assembly always experiences deformations due to internal and external forces/loads and drop in efficiency is critically observed. This paper proposes a methodology to predict such variations in assembly and integrate geometric tolerance design to it. First, the Ontological model of the assembly is predicted through Finite Element Analysis, and the near net shape of the assembly is obtained. Second, a set of features of an assembly which plays a vital role is selected, to determine the optimal geometric tolerances through the hybridised neural network and Genetic algorithm. Finally, a gear pump assembly is chosen, the proposed method is demonstrated. This method will be useful for design and new product development engineers in reducing the assembly variations and associated manufacturing cost.

KEYWORDS:

• Artificial Neural Network (ANN)
• finite element analysis (FEA)
• functional dimensions
• Genetic algorithm (GA)
• Geometric Dimensioning and Tolerancing (GD&T)