An Efficient Model for Batch Annealing Using a Neural Network

A comprehensive process model, capturing heat transfer, phase transformation, and microstructural evolution, was recently developed for efficient process cycle design of a batch annealing operation. However, the high computation time for such an integrated model makes it difficult to design process cycles for a battery of furnaces or their model based online control. In the present work, an accurate prediction tool has been derived using neural network modeling from the simulations of the comprehensive process model. Using this approach, the computational efficiency of this prediction tool increases by over 100 times, making it amenable to cycle design of multiple furnaces or their online process control. Subsequently, an exhaustive search technique and genetic algorithm have been used to optimize the coil dimensions to maximize the plant productivity. It was found that batch annealing productivity can be maximized by optimizing the coil inner diameter, while constraining the coil width and outer diameter at the maximum permissible dimensions.