Neuro - Fuzzy Learning and Genetic Algorithm Approach with Chaos Theory Principles Applying for Diagnostic Problem Solving

A genetic algorithm is a stochastic computational model that seeks the optimal solution to an objective function. A methodology calculation is based on the idea of measuring the increase of fitness and fitness quality evaluation with chaos theory principles applying within genetic algorithm environment. Fuzzy neural networks principles are effectively applied in solved manufacturing problems mostly where multisensor integration, real-timeness, robustness and learning abilities are needed. A modified Mamdani neuro-fuzzy system improves the interpretability of used domain knowledge. The complexity and reliability demands of contemporary industrial systems and technological processes require the development of new fault diagnosis approaches. Due to the large number of process variables and their complex interconnections in a machinery environment, the pertinent knowledge system is mostly qualitative and incomplete. The diagnostic parameters of machine can be represented by a linguistic variable with fuzzy set definition through " fuzzification " in terms of a defined fuzzy membership function. Fuzzy rules describe the qualitative relations between the major operation conditions (i.e. various operation conditions, vibrations, tool accuracy, safe load, working load and so on). In practice, classical tools solve this task by using of sensors great number, or by special sensors creating. This approach increases costs and reduces the reliability. It is possible to compensate this unfavourable state effectively by higher quality of analysis at lower requirements for definiteness of inputs [1]. Chaos theory is a progressive field, which in this time is not sufficiently utilized at failure analysis as possible effective tool. Nowadays, used tools of analysis, or statistic parameters of random phenomena do not find out all possibilities, that is all possible information about real technical state of equipments that the given random signal could provide. Chaos theory principles provide very valuable complementary information, which is not possible to obtain by another means of analysis. Genetic algorithm is a suitable medium for effective application of chaos theory principles for real diagnostic problem solving. It is possible to work with inaccurate, vague, indefinite, non-unambiguous information that real practice in predominant extent provides. Fuzzy neural networks principles are effectively applied in solved manufacturing problems as a first step mostly where multisensor integration, real-timeness, robustness and learning abilities are needed. A modified Mamdani neuro-fuzzy system improves the interpretability of used domain knowledge by parameter data processing. These progressive approaches to complex diagnostic problem solving of practice provide the high effective tools for optimal machine condition and for security of …