Prediction of Humidification Power in Textile-Spinning Mills Using Functional and Neural Networks

The power required for the supply, pumping, and exhaust system of the humidification plant of a spinning department of a textile mill depends on many factors. The most important factors are power of motor driving machinery, lighting and heating load, number of people inside, temperature gradient, and relative humidity. The usual procedure is to train the back-propagation neural network (BPN) with the available data and, once it is trained, BPN will be used for inferring. Some other investigators have used different neural network architectures. In this article, we give a general methodology to build and work with functional network (FN), an alternative to neural network paradigm. In this architecture neural functions, instead of weights, are learned. In addition to data, domain knowledge can also be incorporated. It is shown by means of an example that this functional network architecture is successfully applied to predict power required for the humidification plants.