An adaptive neuro-fuzzy inference system (ANFIS) model for wire-EDM Ulas ß Çaydas ß a, * , Ahmet Hasçalık a , Sami Ekici b a University of Firat, Technical Education Faculty, Department of Manufacturing, Elazig 23 119, Turkey b University of Firat, Technical Education Faculty, Department of Electrical Education, Elazig, Turkey article info Keywords: Wire-EDM Fuzzy inference Artificial neural network Artificial intelligence techniques abstract A wire electrical discharge machined (WEDM) surface is characterized by its roughness and metallo- graphic properties. Surface roughness and white layer thickness (WLT) are the main indicators of quality of a component for WEDM. In this paper an adaptive neuro-fuzzy inference system (ANFIS) model has been developed for the prediction of the white layer thickness (WLT) and the average surface roughness achieved as a function of the process parameters. Pulse duration, open circuit voltage, dielectric flushing pressure and wire feed rate were taken as model’s input features. The model combined modeling function of fuzzy inference with the learning ability of artificial neural network; and a set of rules has been gen- erated directly from the experimental data. The model’s predictions were compared with experimental results for verifying the approach. Ó 2008 Elsevier Ltd. All rights reserved. 1. Introduction Rapid progress in the manufacturing technology has stimulated the application of non-traditional machining (NTM) processes in modern machining to economically machine materials that are usually difficult to machine with the conventional tools (Chakr- aborty & Dey, 2007). Wire electrical discharge machining (WEDM) is one of the most widely applied NTM process for machining and shaping hard, fragile and difficult-cutting in the tool and die indus- try process, which with a thin wire as an electrode transforms elec- trical energy into thermal energy for removing materials (Hasçaly ´k & Çaydas ß, 2004). In WEDM, the erosion mechanism has been de- scribed as melting and/or evaporation of the surface material by the heat generated in the plasma channel. A spark is produced be- tween the wire electrode (usually smaller than 0.3 mm) and work- piece through deionized water, (used as dielectric medium surrounding the workpiece) and erodes the workpiece to produce complex two and three dimensional shapes (Kim & Kruth, 2001). The top surface of the workpiece resolidifies and subsequently cools extremely quickly to form a hard skin on the workpiece. This layer causes an increase in surface roughness and makes the surface hard and brittle. Thus, surface roughness and white layer thickness determine the economics of machining and rate of production. In setting the machining parameters, the main goal is the minimum surface roughness and white layer thick- ness. It is difficult to utilize the optimal functions of a machine owing to there being too many adjustable machining para- meters. Artificial intelligent techniques, such as artificial neural net- works (ANNs) and fuzzy logic, etc., have been successfully applied to machining processes through recent years. A broad literature survey has been conducted on the application of artificial intelli- gence systems to EDM/WEDM. Tarng, Tseng, and Chung (1997) developed a fuzzy pulse discriminator to classify various discharge pulses in EDM. A simulated annealing algorithm was applied to construct the suitable membership function. Trapezoid-shaped membership function was found suitable for the developed fuzzy pulse discriminator which could quickly and accurately classified the discharge pulses under varying cutting conditions. Wang, Gelg- ele, Wang, Yuan, and Fang (2003) developed a hybrid ANN and genetic algorithm (GA) methodology to model and optimize the EDM process. Yilmaz, Eyercioglu, and Gindy (2006) introduced a user-friendly fuzzy-expert system for the selection of the EDM parameters. The fuzzy model and the system were supported by experiments. Fuzzy-expert rules (if–then rules), membership func- tions and defuzzification methods were all used to eliminate the complexity of the situation. As a result, a more precise selection of EDM parameters that are difficult to measure to be taken into consideration were allowed by the developed fuzzy model. Zhang et al. (2002) developed an adaptive fuzzy control system for electro discharge machining with ultrasonic vibration. The discharge pulse parameters and the gap between the tool electrode and the work- piece material were controlled by developed system in a timely manner. Yan and Fang (2007) applied a genetic algorithm based fuzzy logic control in wire transport system of wire-EDM machine. Smooth wire transportation was achieved with no wire breakage during wire feeding. The proposed genetic algorithm based fuzzy 0957-4174/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2008.07.019 * Corresponding author. Tel.: +90 424 2370000/4229; fax: +90 424 2184674. E-mail address: ucaydas@firat.edu.tr (U. Çaydas ß). Expert Systems with Applications 36 (2009) 6135–6139 Contents lists available at ScienceDirect Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa