Abstract:
Due to the wide variability of cotton fibre properties, such as fibre strength , Upper High
Mean Length (UHML), Uniformity Index(UI), Micronaire, Short Fibre % (SF%), Fibre
Elongation, Yellownes, Mean length, Neps, Maturity Ratio(MR) from bale to bale , the
aspect of cotton performance prediction is always very much tricky and arduous job. A large
number of predictive models have been exercised to prognosticate the yarn strength. By and
large, there are three distinguished modeling methods for predicting the yarn properties like
Mathematical models, Statistical regression models and Intelligent models.
A theoretical or Mathematical approach and an empirical or statistical approach both types of
models have their advantages and disadvantages. For instance, the mathematical models are
derived from the first principle analysis and have their basis in applied physics. Therefore,
they are appealing and capable of providing a better understanding of the complex
relationships between the yarn properties and the influencing parameters. However, firstly
these models always require simplified assumptions to make the mathematic tractable, and
the validity of the model depends on the validity of the assumptions. Secondly, the
mathematical models are associated with large prediction errors and therefore not reliable
enough to work in practical situations due to the uncertainties connected with the real world
dynamics. On the other hand, the empirical or statistical models are easy to develop but they
require the specialized knowledge of both statistical methods and designs of experiments.
Extensive experimentation and test and data gathering connected with measurement errors
can generate the `noise' in data. Unfortunately, these models are sensitive to the `noise'. Also
the present techniques are insufficient for precise modeling and predicting the complex nonlinear
processes. The prediction accuracy of ANN has been acclaimed by most of the
researchers. However, ANN modeling has also received criticisms for acting like a ‘black
box’ without revealing much about the mechanics of the process. Some limitation of the
ANN modeling could be overcome by using fuzzy logic, which can effectively translate the
experience of a spinner into a set of expert system rules. It is quite possible to devise a fuzzy
logic based expert system which can predict yarn strength from the given input parameters.
