Concurrent tolerance allocation using an artificial neural network and continuous ant colony optimisation

Abstract

The allocation of tolerances for the components of a mechanical assembly strongly influences manufacturing cost and functional performance. In order to get a reliable tolerances and costs, it is necessary to obtain manufacturing cost-tolerance models. Traditionally, these models are established by various curve-fitting techniques using empirical experimental data. Existing empirical models, however, have considerably large model fitting error, inconsistent modelling accuracy over the tolerance range of typical manufacturing processes. Using these mathematical models will introduce a considerably large error in optimal design of component tolerances. This work presented in this paper uses an artificial neural network (ANN), to overcome above limitations, for establishing manufacturing cost-tolerance models for various manufacturing processes. Having built the ANN cost-tolerance models, continuous ant colony optimisation (CACO) algorithm is used to obtain optimum combination of tolerances for minimum manufacturing cost. A typical tolerance design example is used to illustrate the effectiveness and reliability of the proposed approach.