Optimization of the Process Parameters for Surface Roughness of Work roll Grinding machine using Neural Network

Grinding Process is a complex manufacturing process with a large no. of interacting variables. The rolls used in Sendzimir mills are ground in roll grinding shop to remove the marks formed in the surface of the work rolls during rolling and to remove the oxidizing layer. As Sendzimir mills are friction driven, the work roll must have suitable roughness for thickness reduction. To obtain the average roughness consistently, the process has to be optimized & optimal level of the parameters has to be determined.

In this study, a statistical Taguchi approach & a Back propagation Neural Network Model were devised to evaluate the effects of various parameters & identify the optimal grinding parameters during roll grinding. The Experimental data of surface roughness measured through Taguchi’s Design of Experiment was utilized to train the Back propagation Neural Network model using MATLAB software. The data’s input to MATLAB are trained through Levenberg-Marquardt algorithm. Data’s learns through gradient descent methods.

 Trained Neural Network model was used in predicting the combination of input parameters for the desired surface roughness. The developed prediction system was found to be capable of predicting the accurate combination of input parameters (Wheel speed, Job speed & Traverse speed) for the range it had been trained.

Trained data’s are tested for confirmation of performance. Developed model of ANN is used to predict the combination of input parameters for the expected surface roughness. The same data’s are also analyzed through Taguchi’s DOE. Artificial Neural Network model gives a better result than Taguchi’s Design of Experiments, which could be easily visualized through the result obtained.

Tags :
Your rating: None Average: 3.5 (2 votes)