Pattern Classification Based upon Part-Machine Incidence Matrices

The part-machine grouping problem is individual of the first steps in the manufacturing or design of a cellular manufacturing system. The huge problem occupies the consideration of some real-world complexities as like: part demand volumes, presentence of multiple copies of machines of all type, interchange routings, require balancing workloads in between various cells and also ensuring flexibility.

Numerous neural network techniques have been applied to solve those problems. Mostly of these approaches are based upon unsupervised learning techniques. This is because of the fact that described a set of part routings, patterns of the same routings are not all the time identified completely a priori and, from a practical standpoint, not supervised learning are much more wanted. Unsupervised methods or techniques do not need training and supervised prior learning, and also they also have the ability of processing huge amounts of input data. Unsupervised networks such have been applied can be categorizing as:

(a)    Competitive Learning Model,

(b) Interactive Competition and Activation Model,

(c) Kohonen's Self-Organizing Feature Map model,

(d) Adaptive Resonance Theory Model, and

(e) Fuzzy ART model.