Abstract
For a decade, surface roughness is considered as the main factor of product quality since it has a great influence on the performance of mechanical parts as well as production cost. Surface roughness has an impact on the mechanical properties like fatigue behavior, corrosion resistance, creep life, etc. There are various methods for measurement of surface roughness. They are direct measurement methods, comparison based techniques, non contact methods and on process measurement. The parameters which lead to surface roughness are cutting speed, feed rate, depth of cut, cutting environment, cutting tool wears and so on. In the drilling process, if the speed becomes too high the tool will break and also if the speed becomes too low it will take a lot of time to complete the process and the production rate will go down. Thus, surface finish is an important factor to taken into account in the drilling process. So, it is more necessary to predict the surface roughness of the materials. A lot of researchers have been contributed in predicting the surface roughness of the materials. However, many of them failed since the input model and output categorization varies. Some of the research are ANN model for predicting surface roughness from machining parameters such as cutting speed, feed rate, and depth of cut. Another model is hybrid modeling approach, based on the group method of data handling and the differential evolution population-based algorithm, for modeling and predicting surface roughness in turning operations. But it is difficult to calculate the optimal cutting conditions for the considered material and tool. Also the neural network model coupled with the GA is proposed to determine the optimal machining for surface roughness. But, all these methods fail as there is a large variation in input model and output. Moreover, a recent research was conducted in predicting the surface roughness of materials. This predictive model of surface roughness is created by using back propagation neural network and EM (Electromagnetism) optimization algorithm is used to optimize the problem. The research showed that the EM algorithm coupled with back propagation neural network is an efficient and accurate method in obtaining the minimum of surface roughness. However, in order to further reduce the variation between input model and output, we proposed a feed forward neural network model using APSO (Adaptive particle swarm optimization) algorithm. Our proposed prediction model using APSO algorithm is a very efficient method in decreasing the variation between input model and output than the conventional PSO algorithm. Also, our proposed model minimizes the error to a greater extent than any other method.