genetic algorithms

Engineering materials and structures have crack-like defects leading to premature failures. The usage of fracture mechanics to assess the structural integrity requires knowledge on the type, location, shape, size, and orientation of the flaws. Tomographic reconstruction is one of the commonly used nondestructive testing methods for flaw characterization. The cross sectional image of the object being tested is obtained through the application of various reconstruction methods that are categorized as either analytical methods or iterative methods. In this work, an iterative algorithm that works on the principles of genetic algorithms is developed and used for the reconstruction. The results of simulation studies on the tomographic reconstructions using genetic algorithms for the identification of defects in isotropic materials are discussed in the paper. The solution methodology based on the use of genetic algorithms is applied to reconstruct the cross sections of test specimens with different flaw characteristics. Simulated time-of-flight data of ultrasound rays transmitted through the specimen under investigation is used as input to the algorithm. The time-of-flight data is simulated neglecting the bending of ultrasound rays and assuming straight ray paths. Numerical studies performed on several specimens with flaws of known materials but unknown location, size and shape are presented. The number of ultrasonic transmitters and receivers needed for complete scanning of the specimen’s cross section is analyzed and presented. The findings of the parametric analysis and sensitivity analysis in order to choose appropriate range of algorithm parameters for performance quality and robustness of the algorithm are presented. The performance of the present algorithm with noisy projection data is also discussed.

The placement of wind turbines in a wind farm can considerably affect the total output power. Using computational fluid dynamics and genetic algorithm, the optimal arrangement of turbines in a given wind farm was determined. A three-dimensional Reynolds-averaged Navier-Stokes simulation was conducted on a 660 kW three-bladed horizontal axis turbine. The airflow was assumed to be steady state and a pressure-based approach was adopted to solve the governing equations. By employing the characteristics of the wake propagation, the appropriate distances between the adjacent turbines were calculated. To find the optimal placement of the turbines, a purpose-built genetic algorithm was employed. The results show that the final configuration is in line with the outcomes of the previous study. The sensitivity analysis of the genetic algorithm with respect to its parameters was also performed to guarantee that the final layout is an optimal one.

In this paper, the static pull-in instability (SPIV) of beam-type microelectromechanical systems is theoretically investigated. Herein, modified strain gradient theory in conjunction with Euler–Bernoulli beam theory have been used for mathematical modeling of the size dependent instability of the micro beams. Considering the mid-plane stretching as the source of the nonlinearity in the beam behavior, a nonlinear sizedependent Euler-Bernoulli beam model is used based on a modified couple stress theory, capable of capturing the size effect. Two common beam-type systems including doubleclamped and clamped-free cantilever have been investigated. By selecting a range of geometric parameters such as beam lengths, width, thickness, gaps and size effect, we identify the static pull-in instability voltage. Back propagation artificial neural network (ANN) with three functions have been used for modelling the static pull-in instability voltage of micro beam. Effect of the size dependency on the pull-in performance has been discussed for both micro-structures. The network has four inputs of length, width, gap and the ratio of height to scale parameter of beam as the independent process variables, and the output is static pull-in voltage of microbeam. The number of nodes in the hidden layer, learning ratio and momentum term are optimized using genetic algorithms (GAs). Numerical data, employed for training the network and capabilities of the model in predicting the pull-in instability behaviour has been verified. The output obtained from neural network model is compared with numerical results, and the amount of relative error has been calculated. Based on this verification error, it is shown that the back propagation neural network has the average error of 6.36% in predicting pull-in voltage of cantilever micro-beam. Resultant low relative error value of the ANN model indicates the usability of the BPN in this area.

Tournament selection has been widely used in Genetic Algorithms since it has many advantages such as less time complexity, O(n), applicability for parallel programming, etc. Even if Tournament selection offers more efficiency than other selection methods, it is insufficient to adjust the equilibrium between exploration and exploitation. Because of this insufficiency, it may not convergence to the optimum points closely. The paper introduces a new selection method based on standard Tournament selection. Basically, the new selection method randomly chooses individuals from the population. The method computes fitness values of contestants, then selects the best and the worst individuals between contestants in order to utilize for selection. For some probability value, the selection method decides which individual is chosen, the best or the worst. The new technique overcomes the problem where Tournament is not sufficient by not giving up the exploration. Well-known benchmark functions are used to test the performance of the method. Moreover, the performance of new method is compared with Tournament selection to examine whether or not an improvement occurs. The results show that the new selection technique converges the optimum points closer than the standard Tournament selection.

One of reasons that researchers in recent years have tried to produce ultrafine grained materials is producing lightweight components with high strength and reliability. There are disparate methods for production of ultra-fine grain materials,one of which is severe plastic deformation method. Severe plastic deformation method comprises different processes, one of which is Parallel tubular channel angular pressing. The aim of this study is optimizing parameters of the noticed process die just by utilizing neural network and genetic algorithm methods that at first for this purpose, by using ABAQUS finite element software, the numerical analysis of the die parameters is performed and the impact of each die parameter on the force of the process and the equivalent strain is examined. Finally, for gaining optimal parameters, MATLAB and neural network optimization methods and genetic algorithm are used. The use of neural network and genetic algorithm illustrated that to achieve the ideal possible situation in order to achieve a flawless super-fine tube, it is imperative to use the friction coefficient of 0.05, tube length of 40 mm, channel angle of 140 degrees and diameter increase difference of 1.5 mm. With such values, strain fluctuations reach 0.23, lowest value, and also the force reaches 0.49 KN and the amount of applied strain reaches its highest value to 2.37.

This paper focuses upon the development of an efficient method for conceptual design optimization of a satellite. There are many option for a satellite subsystems that could be choice, as acceptable solution to implement of a space system mission. Every option should be assessment based on the different criteria such as cost, mass, reliability and technology contraint (complexity). In this research, mass and technology constraints, which have a direct impact on the satellite life cycle cost, are considerd as system level objective function to obtain the system optimal solution during the coceptual design phase. The approach adopted in this paper is based on a distributed collaborative optimization (CO) framework. At system level, multiobjective optimization goal is to minimize the dry mass of the satellite and, simultaneously, minimize the system technology complexity which is subject to equality constraints. The use of equality constraints at the system level in CO to represent the disciplinary feasible regions, introduces numerical and computational difficulties as the discipline level optima are non-smooth and noisy functions of the system level optimization parameters.To address these difficulties robust optimization algorithms such as genetic algorithms (GA) are used at the system level. The results show that the CO framework has the same level of accuracy as the conventional All-At-Once approaches.

This paper focuses on the development of an efficient distributed collaborative optimization method for the design of remote sensing small satellite mission in low earth orbit (LEO). The satellite mission requirement involves the duration in which the satellite is able to take images, send data to the ground station and the amount of information it can store. Conventionally, all at once methods are used in satellite mission analysis, however, design optimization of such systems are multidisciplinary task with multiple conflicting objectives such as cost, performance and reliability. The approach adopted in this paper is based on a distributed collaborative optimization (CO) framework. In this approach, the design optimization problem is divided into two levels; namely system and discipline levels. The discipline level optimization involves payload, power, mission and launch subsystems. The objective function of the system level is to minimize the resolution of the satellite imaging payload subject to equality constraints. The use of equality constraints at the system level in CO to represent the disciplinary feasible regions introduces numerical and computational difficulties, as the discipline level optima are non-smooth and noisy functions of the system level optimization parameters. As a result of these difficulties, derivative-based optimization techniques cannot be used for the system level optimization. To address these difficulties a robust optimization algorithm, genetic algorithms (GA), are used at the system level, whilst at the discipline level efficient gradient based techniques are utilized. The results show that distributed CO framework using GA has the same level of accuracy as with the conventional all at once approaches, while providing a potential approach for solving complex multidisciplinary design problems such as the design of satellite systems.