Artificial Neural Networks: Applications to Textiles

摘要

1.1 The Advent of Neural Networks in Textiles 1.2 Overview of Applications in Textiles 1.3 Arrangement of this Monograph The textile process, like many other industrial processes, involves the interaction of a large number of variables. But, unlike most other industries, the relation between the variables and the product properties is yet to be established conclusively. The reasons are the high degree of variability in raw materials, multistage processing, and a lack of precise control on process parameters. The lack of established relationships between raw-material properties, process parameters, and product properties has resulted in a situation in which the manager of a textile mill has to rely primarily on his intuition to keep the process under control. Years of observing the nuances of the process give the manager an uncanny sense about how to react in any situation. Where the absence of complete information about the process prevents the researcher from establishing exact relationships between raw-material properties, process parameters, and product properties, the manager, having absorbed all the data, can make intelligent guesses about the process. It is not necessary, and certainly not possible, for the manager to remember all the details of the process-related data over the years. But, somewhere in his subconscious mind, the information is stacked away, and, on the basis of this vast volume of information, he is able to develop a 'feel' of the process. This is in a way similar to the way a child learns the most mundane tasks like walking, talking, and, in general, interacting with the world. A child takes years to learn the correct way of interacting with the world. He does it, not by being ordered to follow some pre-defined rules, but by looking at the world around him and learning from examples. It is the same principle—'learning from examples'—that the manager adopts, and he is able to exercise some degree of control over the textile process.