Fuzzy control of a system for assembling fibre optic transmitters

This paper presents an automatic alignment system for assembling optical fibre components. A gradient-based algorithm enhanced with momentum information is used to find the location where there is maximum transmitted light from a fibre. The system employs a fuzzy logic controller to optimize the search parameters during the alignment process. The control knowledge base is automatically generated via a new evolutionary algorithm, removing the need for lengthy manual optimization of the control policy. The algorithm evolves a population of fuzzy systems using generational replacement, a new selection operator and a set of crossover and mutation procedures manipulating different elements of the fuzzy knowledge base. The population is divided into three sub-groups to which the different operators are applied. Tested on a batch of fibre optic components, the system achieved 100% success rate with alignment times and accuracies comparing favourably against those obtained using a manually designed controller.