International Journal of Computational Intelligence and Applications

A common step in most of water treatment plants is the chemical coagulation. The chemical coagulation is the process of destabilizing the colloidal particles suspended in raw water by the addition of coagulants. Generally, the determination of the quantity of coagulant to be added to water is made manually by jar tests. However, the manual control has slow response to changes of raw water and it requires intensive laboratory analysis. To reduce the manual effort and to improve the response to change in raw water quality, this work proposes the determination of the coagulant dosage using dynamic neural network modeling using the available sensors as input of the model. The case of study is a large scale water treatment plant in Ceará, Brazil, where the kinds of coagulants added to water are the aluminum sulphate (AS) and poly aluminum chloride (PAC). Several dynamic neural network models with different combinations of the input variables have been evaluated. The best solution found is composed by a nonlinear autoregressive with exogenous input (NARX) model having three input variables, the pH in raw and coagulated water, and the turbidity in the coagulated water, with coefficient of determination of R² = 0.95 and R² = 0.91 for the AS and PAC dosage prediction, respectively.

In this paper, we propose an effective convolutional neural network (CNN) model to the problem of face recognition. The proposed CNN architecture applies fused convolution/subsampling layers that result in a simpler model with fewer network parameters; that is, a smaller number of neurons, trainable parameters, and connections. In addition, it does not require any complex or costly image preprocessing steps that are typical in existing face recognizer systems. In this work, we enhance the stochastic diagonal Levenberg–Marquardt algorithm, a second-order back-propagation algorithm to obtain faster network convergence and better generalization ability. Experimental work completed on the ORL database shows that a recognition accuracy of 100% is achieved, with the network converging within 15 epochs. The average processing time of the proposed CNN face recognition solution, executed on a 2.5 GHz Intel i5 quad-core processor, is 3 s per epoch, with a recognition speed of less than 0.003 s. These results show that the proposed CNN model is a computationally efficient architecture that exhibits faster processing and learning times, and also produces higher recognition accuracy, outperforming other existing work on face recognizers based on neural networks.

With the massive growth and large volume of the web it is very difficult to recover results based on the user preferences. The next generation web architecture, semantic web reduces the burden of the user by performing search based on semantics instead of keywords. Even in the context of semantic technologies optimization problem occurs but rarely considered. In this paper, document clustering is applied to recover relevant documents. We propose an ontology-based clustering algorithm using semantic similarity measure and Fuzzy C-Means, which is applied to the annotated documents for optimizing the result. The proposed method uses Jena API and GATE tool API and the documents can be recovered based on their annotation features and relations. A preliminary experiment comparing the proposed method with K-Means, PSO and hybrid approach PSOK Means shows that the proposed method is feasible and performs better than other clustering methods.

This study presents a novel neuro-fuzzy (NF)-based auto-tuning proportional integral controller (NFATPI) for accurate speed control, and to ensure optimal drive performances of the indirect field controlled induction motor drive, under system disturbances and uncertainties. The training mechanism of the proposed NF have been developed and illustrated through mathematical formulations. Then, the NF parameters have been updated on-line using a suitable training algorithm. The learning rates of the NF are derived on the basis of the discrete Lyapunov function is also illustrated, in order to confirm the stability and the performance of prediction of the proposed NFATPI. The simulation results confirm the effectiveness of the strategy NFATPI as a robust controller for high performance industrial motor drive systems.

In this paper, a novel metaheuristic optimization algorithm inspired by General Relativity Theory (GRT) is presented. In this method that we named General Relativity Search Algorithm (GRSA), a population of particles is considered in a space free from all external non-gravitational fields and propel toward a position with least action. Based on GRT, particles have conserved masses and move along geodesic trajectories in a curved space. Step length and step direction for updating the particles are separately computed using particles velocity and geodesics, respectively. Velocity of particles is obtained by their energy–momentums. According to physical action principle, a population of particles goes to the position with minimum action. By inspiring this physical principle, GRSA will lead variables of an optimization problem move toward the optimal point. Performance of the proposed optimization algorithm is investigated by using several standard test functions and optimal Power System Stabilizers (PSSs) design in a multi-machine power system as a real-world application. Numerical simulations results demonstrate the efficiency, robustness and convergence speed of GRSA in solving various problems.

In this paper, we present a novel Interval-Based Mutation (IBMU) operator. The proposed mutation operator is performing coarse-grained search at initial stage in order to speed up convergence toward more promising regions of the search landscape. Then, more fine-grained search is performed in order to guide the solutions towards the Pareto front. Computational experiments indicate that the proposed mutation operator performs better than conventional approaches for solving several well-known benchmarking problems.

In this paper, an Interactively Recurrent Self-evolving Fuzzy Cerebellar Model Articulation Controller (IRSFCMAC) model is developed for solving classification problems. The proposed IRSFCMAC classifier consists of internal feedback and external loops, which are generated by the hypercube cell firing strength to itself and other hypercube cells. The learning process of the IRSFCMAC gets started with an empty hypercube base, and then all of hypercube cells are generated and learned online via structure and parameter learning, respectively. The structure learning algorithm is based on the degree measure to determine the number of hypercube cells. The parameter learning algorithm, based on the gradient descent method, adjusts the shapes of the membership functions and the corresponding fuzzy weights of the IRSFCMAC. Finally, the proposed IRSFCMAC model is tested by four benchmark classification problems. Experimental results show that the proposed IRSFCMAC model has superior performance than traditional FCMAC and other models.