Abstract
This paper presents an approach for the development of surrogate models predicting the assembly time of a system based on complexity metrics of the physical system architecture when detailed geometric information is unavailable. A convention for modelling physical architecture is presented, followed by a sample of 10 analysed systems used for training and three systems used for validation. These systems are evaluated on complexity metrics developed from graph theoretic measures. An example model is developed based on a series of regressions of trends observed within the sample data. This is validated against the systems that are not used to develop the model. The model developed uses average path length, part count and path length density to approximate assembly time within the standard deviation of the subjective variation possible in Boothroyd and Dewhurst design for assembly (DFA) analysis. While the specific example model developed is generalisable only to systems similar to those in the sample set, the capability to develop mappings between physical architecture and assembly time in early-stage design is demonstrated.
Keywords:
Acknowledgements
The authors would like to recognise the efforts of the students and members of our research group which helped provide us with source data and those figures for which a specific source was not cited.