LITERATURE REVIEW
2.3 Analysis of different methodologies for crack detection
2.3.3 Crack detection using AI technique
2.3.3.3 Genetic algorithm method
In the process of development of various methods for crack identification genetic algorithm is also used efficiently for accurate measurement of the damage location and depth and also fault detection in engineering systems. The genetic algorithm based methodologies are discussed in this section.
Meruane et al. [110] have implemented an hybrid real-coded Genetic Algorithm with damage penalization to locate and quantify structural damage. The performance of five fundamental functions based on modal data is studied by them. In addition, the authors have proposed the use of a damage penalization that satisfactorily avoids false damage detection due to experimental noise or numerical errors. They have tested the effectiveness of the proposed technique on a tridimensional space frame structure with single and multiple damages scenarios and stated that this approach reaches a much more precise solution than conventional optimization methods. Nobahari et al. [111] have proposed an efficient optimization procedure using genetic algorithm to detect multiple damage in structural systems based on the changes in the natural frequency. They have applied finite element analysis to evaluate the required natural frequencies. Two numbers of bench mark tests have been utilized to demonstrate the computational advantages of the proposed method by them. Li et al. [112] have presented a novel feature extraction and selection scheme for hybrid fault diagnosis of gearbox based on transform function, non-negative matrix factorization (NMF) and multi-objective evolutionary genetic algorithms. The transform function has been adapted to acquire the vibration signals for various fault condition of the gear system and the
en– e-n a =
en+ e-n
non-negative matrix factorization (NMF) was employed to extract features from the time– frequency representations. The genetic algorithm has been used for accurate classification of hybrid faults of gearbox. Results from the experiments as described by them revealed that the proposed feature extraction and selection scheme demonstrate to be an effective and efficient tool for hybrid fault diagnosis of gearbox. Fernando et al. [113] have dealt with the crack detection in structural elements by means of a genetic algorithm optimization method taking into account the existence of contact between the interfaces of the crack. They have addressed bi- and three-dimensional models to handle the dynamics of a structural element with a transverse breathing crack. Physical experiments have been performed by them with a cantilever damaged beam and the resulting data are used as input in the fault diagnostic genetic algorithm. The benefits of applying automated fault detection and diagnosis to chillers include less expensive repairs, timely maintenance, and shorter downtimes. Han et al. [114] have employed feature selection (FS) techniques, such as mutual-information-based filter and genetic algorithm to help search for the important sensors in data driven chiller fault detection and diagnosis applications, to enhance the performance of fault identification technique. The results shows that the eight features/sensors, centered around the core refrigeration cycle and selected by the proposed method, outperform the other three feature subsets by the linear discriminant analysis. Hussain et al. [115] have described a novel method for real time fault detection in gearboxes using adaptive features extraction algorithm to deal with non-stationary faulty signals. They have claimed that their proposed method is based on combination of conventional one-dimensional and multi-dimensional search methods, which showed high performance and accurate fault detection results compared with evolutionary algorithms like genetic algorithms. Singh et al. [116] have developed a two stage identification methodology, which identifies a number of cracks, their locations on a cracked shaft and its sizes. In the methodology they have utilized transverse forced responses of the shaft system at different frequencies of a harmonic excitation. A multi-objective genetic algorithm technique has been designed using the frequency response of the dynamic structure for crack detection in shaft like structures. Lei et al. [117] have proposed a new multidimensional hybrid intelligent diagnosis method to identify different categories and levels of gear damage automatically using Hilbert transform, wavelet packet transform (WPT) and empirical mode decomposition (EMD) methods to extract additional fault
characteristic information. They have used the extracted features of the system to develop the multidimensional features based genetic algorithm technique to identify gear faults. Sette et al. [118] have presented a method to simulate a complex production process using a neural network and the optimization by genetic algorithm for quality control of the end product in a manufacturing environment. He has applied this method to a spinning production process where input parameters are machine settings and fiber quality, and the yarn strength, elongation are output parameters for the neural network model. He has used the genetic algorithm with a sharing function and a Pareto optimization to optimize the input parameters for obtaining the best yarns. According to him the results from this method are considerably better than current manual machine intervention. Xiang et al. [119] have proposed a new method for crack location and depth in a shaft by following rotating Rayleigh-Euler and Rayleigh-Timoshenko beam elements of B-spline wavelet on the interval. He has described that the cracked shaft is modeled by using wavelet-based elements to gain precise frequencies. According to him the 1st three frequencies are measured to locate the crack and the depths are detected by genetic algorithm. The robustness of the proposed method has been validated by some numerical examples and experimental cases and he has concluded that the method is capable of the detecting the crack in a shaft. He et al. [120] have studied the crack detection in a rotating machine shaft by using finite element method to optimize the problem and subsequently used genetic algorithm to search the solution. Their proposed method has been found to solve a wide range of inverse identification problem. Zhang et al. [121] have used genetic programming (GP) in finding faults in rotating machinery. They compared the solution through GP with other techniques like artificial neural network (ANN) and support vector machines (SVMs). They have found that GP demonstrates performance equal or better compared to ANN and SVMs. Zhang et al. [122] have studied the fault in rolling element bearing by the combination of genetic algorithm (GA) and fast kurtogram. For the initial analysis of the vibration signals of the bearing they have used fast kurtogram and subsequently for final optimization they have used GA The results of their combined applications of GA and kurtogram have been found to give better results over the other optimal resonance demodulation techniques. Baghmisheh et al. [123] have used genetic algorithm (GA) to monitor the changes in natural frequencies of a cantilever beam having crack. They have used an analytical model to formulate the crack beam structure and
numerical methods to obtain the natural frequencies. The depths and crack locations have been solved by using binary and continuous genetic algorithms BGA, CGA). Perera et al. [124] have used genetic algorithm for solving multi objective optimization to detect damage. They have compared GA optimizations based on aggregating functions with pareto optimality. Friswell et al. [125] have combined genetic algorithm (GA) and eigen sensitivity method for determination of location of damage in structures. The GA has been used by them to optimize the discrete damage location variables. They have used eigen sensitivity method to optimize the damage extent.