Structural Health Monitoring Using Genetic Fuzzy Systems

Chapter 1 defines the various terms and definitions related to structural health monitoring. The main technical approaches used for structural health monitoring are discussed. These include methods based on mathematical models of the undamaged and damaged structures and those based on modal data such as frequencies and mode shapes. In addition, localized damage detection methods based on strain monitoring and non-destructive testing are described. Three main soft computing methods (neural networks, genetic algorithms, and fuzzy logic) used for solving the structural health monitoring problem are discussed with reference to the published literature. The advantages and shortcomings of each of these methods are brought out, and techniques which hybridize these methods are shown to be good alternatives. The genetic fuzzy system is introduced as an excellent choice for solving the pattern recognition problems in structural health monitoring. The helicopter rotor system health and usage monitoring system is used as an example to illustrate the different ideas for a real engineering system. The chapter ends with a summary of the book.

Chapter 2 gives a brief introduction of fuzzy logic, genetic algorithms, and several other concepts used in the formulation of the genetic fuzzy system. Concepts such as fuzzy sets, fuzzy rule base, inference engine, linguistic measure, universe of discourse, fuzzification, and defuzzification are explained. The concept of converting numbers into words using fuzzy membership functions, which is critical in fuzzy logic, is illustrated. The genetic algorithm is introduced as an optimization method used for maximizing an objective function. The genetic algorithm is based on the survival of the fittest theory from evolution. The selection of a starting population and a mating pool is explained. The genetic operators of crossover and mutation are illustrated with examples. The use of the genetic algorithm to create optimal fuzzy logic systems by maximizing the success rate of fault isolation is explained. The genetic fuzzy system is the algorithmic tool which is used for fault isolation in this book.

Chapter 3 illustrates the genetic fuzzy system for structural health monitoring of a beam. The cantilever beam structure is fixed at one end and free at the other and can model airplane wings, helicopter blades, and turbine blades, among other things. A finite element model of the undamaged and damaged beam is used to simulate the damaged system. Damage is introduced in the beam by a localized stiffness reduction in accordance with continuum mechanics theory. A genetic fuzzy system is formulated for the beam structure to find the damage location and size from modal data. Initially, the genetic fuzzy system is developed for a uniform beam. It is shown that the method is robust to noise in the data and to missing measurements. The genetic fuzzy system is then illustrated for a nonuniform beam and for a more refined discretization of the damage locations. It is shown that the architecture of the genetic fuzzy system allows for a simple approach for recreating the pattern recognition algorithms as the measurements and outputs change. For each input-output set, a genetic algorithm is used to maximize the success rate of fault isolation for a set of noisy data. The approach is finally illustrated for a non-rotating helicopter main rotor blade using frequency and mode shape data. It is found that the genetic fuzzy system performs very well for the structural health monitoring of beams.

Chapter 4 develops a genetic fuzzy system for structural health monitoring of thin-walled composite tubes. Such tubes are used for power transmission poles and in other important structures. The vibration analysis of the composite tube is conducted using the finite element method. A long pole proposed for power transmission is used for the numerical simulations. Matrix crack damage in the composite material is modeled using a theory based on effective elastic modulus applicable to cross-ply laminates. Composite materials are susceptible to many types of damage, but matrix crack is typically the first damage which occurs in these materials. Furthermore, matrix crack saturation marks the onset of other more serious damages such as delamination and fiber breakage. The genetic fuzzy system is developed to detect the location and size of matrix cracks in the composite pole. The genetic fuzzy system is developed by maximizing the success rate for damage detection under noisy conditions and is tested with noisy data. The misclassification behavior of the fuzzy system is analyzed, and it is found that the genetic fuzzy system works very well for the health monitoring problem.

Chapter 5 illustrates the genetic fuzzy system for health monitoring of a composite helicopter rotor in forward flight. The rotor is the most important component of the helicopter, and its health is critical for helicopter performance and control. Progressive damage accumulation is considered in the composite material. This damage model considers matrix cracking as the first damage type, followed by debonding/delamination, and finally fiber breakage. The damaged helicopter rotor is modeled using a finite element simulation which solves the rotor blade equations and vehicle trim equations. This aeroelastic simulation provides the blade response, blade and hub loads, strains, etc., for a damaged composite helicopter rotor in forward flight. The genetic fuzzy system is developed and tested for this helicopter rotor health monitoring problem. Different combinations of measurements are considered, and their advantages and shortcomings are evaluated. Finally, a life prediction approach is developed based on phenomenological damage growth models, and the genetic fuzzy system is illustrated for damage detection as well as life prediction for a helicopter rotor.