No abstract

An artificial neural network model of 'Feed Forward Back Propagation' type is developed for the analysis and prediction of surface roughness, the relationship between cutting and process parameters of Al-4.5Cu-1.5TiC Metal Matrix Composites. The effect of the process parameters namely, Cutting speed, feed, depth of cut upon the responses like: surface roughness parameter Ra, Rz and Rt of Al-4.5Cu-1.5TiC MMC are analyzed during this investigation. The Experiments have been carried out as per Taguchi's L25 orthogonal array with five levels defined for each of the factors for developing the knowledge base for ANN training. To have all the data in a same scale the experimental results have been normalized before being used in the Artificial Neural Network model.

In modern in situ composite fabrication processes, the selection of optimal process parameters is greatly important for the preparation of best quality metal matrix composite. For achieving high-quality composite, an efficient optimization technique is essential. The present study explores the potential of a new robust algorithm named teaching-learningbased optimization algorithm for in situ process parameter optimization problems in fabrication of Al-4.5%Cu-TiC metal matrix composite fabricated by stir casting technique. Optimization process is carried out for optimizing the in situ processing parameters i.e. pouring temperature, stirring speed, reaction time for achieving better mechanical properties, i.e. better microhardness, toughness, and ultimate tensile strength. Taguchi's L 25 orthogonal array design of experiment was used for performing the experiments. Grey relational analysis is used for the conversion of the multiobjective function into a single objective function, which is being used as the objective function in the teaching-learning-based optimization algorithm. Confirmation test results show that the developed teaching-learning-based optimization model is a very efficient and robust approach for engineering materials process parameter optimization problems.

No abstract

No abstract