The present study has been taken up to emphasize the role of the hybridization process for optimizing a given reinforced concrete (RC) frame. Although various primary techniques have been hybrid in the past with varying degree of success, the effect of hybridization of enhanced versions of standard optimization techniques has found little attention. The focus of the current study is to see if it is possible to maintain and carry the positive effects of enhanced versions of two different techniques while using their hybrid algorithms. For this purpose, enhanced versions of standard particle swarm optimization (PSO) and a standard gravitational search algorithm (GSA), were considered for optimizing an RC frame. The enhanced version of PSO involves its democratization by considering all good and bad experiences of the particles, whereas the enhanced version of the GSA is made self-adaptive by considering a specific range for certain parameters, like the gravitational constant and a set of agents with the best fitness values. The optimization process, being iterative in nature, has been coded in C++. The analysis and design procedure is based on the specifications of Indian codes. Two distinct advantages of enhanced versions of standard PSO and GSA, namely, better capability to escape from local optima and a faster convergence rate, have been tested for the hybrid algorithm. The entire formulation for optimal cost design of a frame includes the cost of beams and columns. The variables of each element of structural frame have been considered as continuous and rounded off appropriately to consider practical limitations. An example has also been considered to emphasize the validity of this optimum design procedure.
In this paper, a new Multi-Layer Perceptron Neural Network (MLP NN) classifier is proposed for classifying sonar targets and non-targets from the acoustic backscattered signals. Besides the capabilities of MLP NNs, it uses Back Propagation (BP) and Gradient Descent (GD) for training; therefore, MLP NNs face with not only impertinent classification accuracy but also getting stuck in local minima as well as lowconvergence speed. To lift defections, this study uses Adaptive Best Mass Gravitational Search Algorithm (ABGSA) to train MLP NN. This algorithm develops marginal disadvantage of the GSA using the bestcollected masses within iterations and expediting exploitation phase. To test the proposed classifier, this algorithm along with the GSA, GD, GA, PSO and compound method (PSOGSA) via three datasets in various dimensions will be assessed. Assessed metrics include convergence speed, fail probability in local minimum and classification accuracy. Finally, as a practical application assumed network classifies sonar dataset. This dataset consists of the backscattered echoes from six different objects: four targets and two non-targets. Results indicate that the new classifier proposes better output in terms of aforementioned criteria than whole proposed benchmarks.
Due to its unique features, the metal foam is considered as one of the newest acoustic absorbents. It is a navel approach determining the structural properties of sound absorbent to predict its acoustical behavior. Unfortunately, direct measurements of these parameters are often difficult. Currently, there have been acoustic models showing the relationship between absorbent morphology and sound absorption coefficient (SAC). By optimizing the effective parameters on the SAC, the maximum SAC at each frequency can be obtained. In this study, using the Benchmarking method, the model presented by Lu was validated in MATLAB coding software. Then, the local search algorithm (LSA) method was used to optimize the metal foam morphology parameters. The optimized parameters had three factors, including porosity, pore size, and metal foam pore opening size. The optimization was applied to a broad band of frequency ranging from 500 to 8000 Hz. The predicted values were in accordance with benchmark data resulted from Lu model. The optimal range of the parameters including porosity of 50 to 95%, pore size of 0.09 to 4.55 mm, and pore opening size of 0.06 to 0.4 mm were applied to obtain the highest SAC for the frequency range of 500 to 800 Hz. The optimal amount of pore opening size was 0.1 mm in most frequencies to have the highest SAC. It was concluded that the proposed method of the LSA could optimize the parameters affecting the SAC according to the Lu model. The presented method can be a reliable guide for optimizing microstructure parameters of metal foam to increase the SAC at any frequency and can be used to make optimized metal foam.