Pensee Journal Vol 75, No. 11;Nov 2013 159 office@penseejournal.com Estimation of ANN Modelling of Laser Cutting with Missing Values Mustafa Zaidi (Corresponding author) Department of Computing, Shaheed Zulfikar Ali Bhutto Institute of Science & Technology (SZABIST) 90 Clifton, Karachi, Pakistan Tel: 00923005269252 e-mail: mustafainisb@gmail.com Imran Amin Department of Computing, Shaheed Zulfikar Ali Bhutto Institute of Science & Technology (SZABIST) 90 Clifton, Karachi, Pakistan e-mail: imran.amin@szabist.edu.pk and amin.imran@gmail.com Ahmad Hussain Department of Nuclear Engineering, King Abdulaziz University, Jeddah, Kingdom of Saudi Arabia e-mail: ahassain@kau.edu.sa Nukman Yusoff 2 Manufacturing Systems Integration, Department of Mechanical Engineering, University of Malaya, Malaysia e-mail: nukman@um.edu.my Abstract Orthogonal array based experimental data were trained by Artificial Neural Network (ANN) for the modelling of laser cutting process of Perspex sheet. The simulation results were compared with factorial design experimental observations. The factorial design is selected for estimation of error on each dataset of observation table with missing values. The ANN simulation evaluation results are expected to depict better generalization based on very small training datasets by applying feed forward back-propagation. The study shows the benefits and disadvantages of neural network parameters variation. The increment in size of training datasets based on factorial design generate better accuracy in simulations as compared to orthogonal array modelling. The research concludes that ANN modelling can be utilized by the new researcher without much trouble from the missing values in observation tables. Keywords: Training algorithm, missing values, Kerf width, edge quality, laser cutting 1. Introduction Laser cutting is a thermal cutting process which is one of the latest technologies in machining materials (Choudhury & Shirley, 2010). The use of laser technology is justified as the process is reliable and produces better quality products, even though it has a high cost, which is constantly reducing (Mustafa & Amin, et al. 2010). Utilization for cutting of plastic materials increases to achieve a finer product quality, together with robust process solution. The motivation of this study is to encourage new researcher/engineers to utilize their expansive machines with the adjustment of small size experimentation. Newcomers’ experiments observation tables often contain missing values during the inappropriate range adjustment and other possible reasons. The experiment was conducted on low cost Perspex material but the method can be utilized on other materials. The CO 2 laser cutting machine adjusted with four important parameters (the speed at which the laser nozzle moves, the distance between sheet and nozzle, laser power and air pressure) which effect output on edge quality, kerf width, overcut and material removal rate. Percent overcut and material removal rate can be calculated from kerf width data therefore Edge quality and kerf widths simulation is sufficient. The problem of modelling can be resolved by the existing