Power Consumption Predictive Tool for Grinding Process Based on Anfis Method

Grinding is considered nowadays as a very important manufacturing process, in terms of both industrial and experimental applications, due to its capability of producing high quality workpiece surfaces. This process is of particular interest in achieving accurate dimensions of parts, machining hard materials that are difficult to be easily processed with other machine tools and is even employed when it is required to remove a large bulk of material rapidly. The grinding wheel constitutes the cutting tool for the grinding process; numerous abrasive particles, called abrasive grains, exist on its surface of the grinding wheel and function as singlepoint cutting edges. Some characteristics of the grinding process are the relatively shallow depth of cut and the extremely small dimension of chip. Furthermore, another intrinsic characteristic of grinding process is the significant amount of energy per unit volume of removed material required for this machining process, as it is calculated by the grinding forces during the process. Apart from direct measurements and monitoring through specialized systems, analytical, numerical and soft computing methods have already been developed and used as valuable tools to study various aspects of machining processes and among them, power consumption. These tools are shown to provide engineers detailed insight of the underlying mechanisms of machining processes and also allow for the determination of contribution of each machining parameter on the outcome of process, usually without considerable cost. Thus, especially soft computing methods are increasingly employed in relevant applications. In the current paper, a soft computing model, for the prediction of energy consumption in a grinding process in relevance to process parameters such as grinding wheel type, workpiece material and depth of cut using a Sugeno-type ANFIS model is developed. The ANFIS model will be applied to a real machining process case and the results of prediction will be discussed in order to account for the efficiency of the proposed model.