Manufacturing Assembly Time Estimation Using Structural Complexity Metric Trained Artificial Neural Networks

Abstract

Assembly time estimation is traditionally a time-intensive manual process that requires detailed geometric and process information, which is often subjective and qualitative in nature. As a result, assembly time estimation is rarely applied during early design iterations. In this paper, the authors explore the possibility of automating the assembly time estimation process while reducing the level of design detail required. In this approach, they train artificial neural networks (ANNs) to estimate the assembly times of vehicle subassemblies using either assembly connectivity or liaison graph properties, respectively, as input data. The effectiveness of estimation is evaluated based on the distribution of estimates provided by a population of ANNs trained on the same input data using varying initial conditions. Results indicate that this method can provide time estimates of an assembly process with ±15% error while relying exclusively on the geometric part information rather than process instructions.