This paper aims at describing the work conducted at Textile Engineering Department, University of Minho (Portugal) in the field of the applications of neural and fuzzy systems on Cotton Spinning.

Those days, the field has gained a more solid background by linking to the traditional systems sciences.

A textile engineer, who is faced with the characterization or the prediction of the plant behavior, has to model the considered process. The needs for process models arise from various requirements. In process design, one wants to understand the mechanical and physical phenomena in order to develop the process.

In control, the short-term behavior and dynamics of the process may need to be predicted. Anomalies in different parts of the process can be detected by comparing a model of known behavior with the measured behavior. The optimal operating strategies can be examined by simulating the process behavior under different conditions.

For linear processes, a multitude of efficient techniques exist already, as linear regression can be used in identification. For simple input-output relations, linear models are a relatively robust alternative. They are simple and efficient also when extending to the identification of adaptive dynamic models, and readily available control design methods can be found from the literature. With suitable preprocessing or reparametrisation, a seemingly non-linear problem may often be converted into a linear one.

However, cotton spinning processes are non-linear and poorly known. As the processes become more complex, a sufficiently correct non-linear input-output behavior is more difficult to obtain using linear methods. Whereas the linear black-box models have been extensively studied and can be handled fairly well, the non-linear case is more difficult. The literature is spread under various fields, such as neural networks and fuzzy models.

Our current emphasis is in neuro-fuzzy systems, where we expect to find the way to create models so transparent, that even a less experienced textile engineer faced with the need of characterization of a spinning plant can find them useful.

Neuro-fuzzy combination is considered for several years already. However, the term "neuro-fuzzy" still lacks of proper definition, and it has the flavor of a "buzz word". In this paper we try to give it a meaning in the context of fuzzy classification systems. From our point of view "neuro-fuzzy" means the employment of heuristic learning strategies derived from the domain of neural network theory to support the development of a fuzzy system. We illustrate our ideas using our "TEXPERT NEUROFUZZY CLASSER" model, which is used to create a fuzzy classification system from data.