State-of-the-art in empirical modelling of rapid prototyping processes A. Garg and K. Tai School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore, and M.M. Savalani Department of Industrial Systems and Engineering, The Hong Kong Polytechnic University, Hong Kong Abstract Purpose – The empirical modelling of major rapid prototyping (RP) processes such as fused deposition modelling (FDM), selective laser sintering (SLS) and stereolithography (SL) has attracted the attention of researchers in view of their contribution to the overall cost of the product. Empirical modelling techniques such as artificial neural network (ANN) and regression analysis have been paid considerable attention. In this paper, a powerful modelling technique using genetic programming (GP) for modelling the FDM process is introduced and the issues related to the empirical modelling of RP processes are discussed. The present work aims to investigate the performance of various potential empirical modelling techniques so that the choice of an appropriate modelling technique for a given RP process can be made. The paper aims to discuss these issues. Design/methodology/approach – Apart from the study of applications of empirical modelling techniques on RP processes, a multigene GP is applied to predict the compressive strength of a FDM part based on five given input process parameters. The parameter setting for GP is determined using trial and experimental runs. The performance of the GP model is compared to those of neural networks and regression analysis. Findings – The GP approach provides a model in the form of a mathematical equation reflecting the relationship between the compressive strength and five given input parameters. The performance of ANN is found to be better than those of GP and regression, showing the effectiveness of ANN in predicting the performance characteristics of the FDM part. The GP is able to identify the significant input parameters that comply with those of an earlier study. The distinct advantages of GP as compared to ANN and regression are highlighted. Several vital issues related to the empirical modelling of RP processes are also highlighted in the end. Originality/value – For the first time, a review of the application of empirical modelling techniques on RP processes is undertaken and a new GP method for modelling the FDM process is introduced. The performance of potential empirical modelling techniques for modelling RP processes is evaluated. This is an important step in modernising the era of empirical modelling of RP processes. Keywords Selective laser sintering, Rapid prototyping, Artificial neural network, Empirical modelling, Fused deposition modeling, Stereolithograhpy Paper type Research paper 1. Introduction Rapid prototyping (RP) is one of the most promising technologies widely used today to reduce product development time by way of realising the prototype component that can be directly used in assemblies, product testing or tooling for short or medium run production. The focus of industries has shifted from the traditional product development methodology to the rapid fabrication techniques. RP relates to a rapidly growing number of automated machines or processes like stereolithography (SL), selective laser sintering (SLS), fused deposition modelling (FDM), laminated object manufacturing (LOM) and shape deposition manufacturing (SDM) which can fabricate three dimensional (3D) solid models from the CAD data automatically without use of tooling and human intervention (Bernard and Fischer, 2002; Hopkinson et al., 2006; Mansour and Hague, 2003; Pham and Gault, 1998; Upcraft and Fletcher, 2003; Wiedemann and Jantzen, 1999; Yan and Gu, 1996). This has resulted in applications in functional prototype development, medical, automobile industries, construction industries, space applications, tool and die making (Sood et al., 2011a). RP processes disperse 3D CAD models into a series of two-and-a-half dimensional slice models with corresponding software. The 3D model is further converted into a series of two and a half dimensional layers. In spite of having such advantages, the full scale application of RP processes has not gained much attention due to compatibility of presently available materials with RP processes. To overcome this problem, a possible route may be to develop new materials with superior characteristics compared to the conventional ones. Another possible route may be to adjust the RP process parameters during the fabrication stage so that the performance characteristics (output) of the fabricated prototype such as surface quality, part strength, build time, accuracy and repeatability improve. Studies by researchers reveal that these characteristics are related to the process parameters and can be improved with proper adjustment without incurring additional expenses for changing hardware and software (Sood et al., 2011a). The cost of a fabricated prototype from such processes is reasonably high. Therefore, from a technological and economic point of view, the process The current issue and full text archive of this journal is available at www.emeraldinsight.com/1355-2546.htm Rapid Prototyping Journal 20/2 (2014) 164–178 q Emerald Group Publishing Limited [ISSN 1355-2546] [DOI 10.1108/RPJ-08-2012-0072] Received: 8 March 2012 Revised: 5 February 2013 Accepted: 4 March 2013 164