|Ph.D Student||Manor Erez|
|Subject||Engineering Fuzziness in Quality and Reliability|
|Department||Department of Quality Assurance and Reliability||Supervisor||Professor Emeritus Amos Notea|
|Full Thesis text|
Quality and reliability domains of interest rely on accurate perception of reality, and the ability to continuously make informed decisions. Fuzziness is an intrinsic property of any material, process, system, and measurement. Fuzziness arises from our inaccurate perception of reality, the fuzzy nature of materials and phenomena, and even our decision making models may require fusion of different knowledge bodies.
Fuzzy Logic is a framework developed to allow imprecise computing. The works pioneered by Zadeh, Mamdani, Sugeno and others, have been used successfully to create models of human knowledge and reasoning applied to controllers. Considerable research was published on the mathematical aspects of fuzzy sets theory, but surprisingly the design of the fuzzy system is not based on scientific or engineering principles, but rather arbitrary and subjective.
In the present research a new methodology was developed to overcome the engineering's void in the existing fuzzy framework by introducing scientific principles into the fuzzy logic framework, and use fuzzy principles to describe and analyze complex physical phenomena.
Firstly, fuzziness was defined as the blurring of crisp domains, through fuzzification functions or operators. The fuzzification is based on scientific principles related to the properties of the fuzzified entities and physical fuzzification processes, and also on engineering criteria such as cost-based optimization. The fuzzification was demonstrated on one and two-dimensional domains, taking into account the correlation between multiple inputs.
The fuzzy principle was applied also to physical laws, for materials or systems exhibiting partial membership to any of several distinct physical modes. A fuzzy fusion of such distinct laws creates a new law that describes dynamically the interim state of the material or system.
The methodology was applied to selected problems from the quality and reliability engineering fields, exhibiting fuzzy properties. Nondestructive radiograph blurred images were described by PSF fuzzification functions. A classification problem of gas mixtures, through cost-based two-dimensional fuzzification, increased the accuracy. A fuzzification process and measurement (de-fuzzification) method of nano-particles with designed fuzzy hydrophobicity was demonstrated in a series of experiments. The visco-elastic phenomenon was modeled by fuzzifying the viscous and the elastic models into a new model. Comparing empirical data with models was performed with a dynamic integro-differential model, with partial derivatives models, and a new operator evolving from the latter. Lastly, a fuzzification - de-fuzzification tradeoff was shown to increase the reliability and strength accuracy of carbon fibers, while increasing the fuzziness of the internal state of the fiber.