Control chart pattern identification using a synergy between neural networks and bees algorithm

Control chart pattern recognition using RBF neural network with new training algorithm and practical features

Control charts are employed in manufacturing industry for statistical process control (SPC). It is possible to detect incipient problems and prevent a process from going out of control by identifying the type of patterns displayed by the control charts. Various techniques have been applied to this control chart pattern recognition task. This paper presents the use of learning vector quantisation (LVQ) networks for recognising patterns in control charts. The LVQ networks were trained, not by applying standard training algorithms, but by employing a new optimisation algorithm developed by the authors. The algorithm, called the bees algorithm, is inspired by the food foraging behaviour of honey bees. The paper first describes the bees algorithm and explains how the algorithm is employed to train LVQ networks. It then discusses the recognition of control chart patterns by LVQ networks optimised using the bees algorithm

In this study, an integrated multi-layer perceptron neural network and Bees Algorithm is presented for analyzing world CO2 emissions. For this purpose, the following steps are done: STEP 1: In the first step, the Bees Algorithm is applied in order to determine the world's fossil fuels and primary energy demand equations based on socio-economic indicators. The world's population, gross domestic product, oil trade movement, and natural gas trade movement are used as socio-economic indicators in this study. The following scenarios are designed for forecasting each socio-economic indicator in a future time domain: Scenario I: For each socio-economic indicator, several polynomial trend lines are fitted to the observed data and the best fitted polynomial (highest correlation coefficient (R2) value) for each socio-economic indicator is used for future forecasting. Scenario II: For each socio-economic indicator, several neural networks are trained and the best trained network for each socio-economic indicator is used for future forecasting. STEP 2: In the second step, world CO2 emission is projected based on the oil, natural gas, coal, and primary energy consumption using Bees Algorithm. The related data from 1980 to 2006 are used, partly for installing the models (1980–1999) and partly for testing the models (2000–2006). World CO2 emission is forecasted up to year 2040.

This paper proves the capability of the bees algorithm to solve complex parameter optimization problems for robot manipulator control. Two applications are presented. The first case considers the modelling of the inverse kinematics of an articulated robot arm using neural networks. The weights of the connections between the nodes need to be set so as to minimize the difference between the neural network model and the desired behaviour. In the proposed example, the bees algorithm is used to train three multilayer perceptrons to learn the inverse kinematics of the joints of a three-link manipulator. The second case considers the design of a hierarchical proportional–integral–derivative (PID) controller for a flexible single-link robot manipulator. The six gains of the PID controller need to be optimized so as to minimize positional inaccuracies and vibrations. Experimental tests demonstrated the validity of the proposed approach. In the first case, the bees algorithm proved very effective at optimizing the neural network models. Compared with the results obtained employing the standard back-propagation rule and an evolutionary algorithm, the bees algorithm obtained superior results in terms of training accuracy and robustness. In the second case, the proposed method demonstrated remarkable efficiency and consistency in the tuning of the PID controller parameters. In 50 independent optimization trials, the PID controllers designed using the bees algorithm consistently outperformed a robot controller designed using a standard manual technique.

The rainy season is only available in tropical climates. One of the problem that occurs during the rainy season is flooding. Hence forecasting is needed to predict rainfall. This research proposes a rainfall forecasting using backpropagation method with Bee colony optimization. One problem with the backpropagation method is the initialization of initial weights and random biases. Where the backpropagation weight value can be optimized using the bee colony algorithm. The initial weight can be optimized by the bee colony method. The bee colony is included in a swarm intelligence that can be used to optimize weight and bias form backpropagation method. Expected results from the backpropagation method could be improved with bee colony optimization. Results of the study shows that the learning rate of 0.3 MSE with 0.0111 results. From the previous results with the best learning rate of 0.3, then optimization of weight and bias with bee colony algorithm. Many foods sources and the number of bees as much as 3 MSE with 0.00939 results. Therefore, it is concluded that bee colony can optimize the weight and biases of backpropagation to get minimal errors with learning $\mathbf{rate} =3$ and number of $\mathbf{bees} =3$ .

In this paper, the Bees algorithm was used to train multi-layer perceptron neural networks to model the inverse kinematics of an articulated robot manipulator arm. The Bees Algorithm is a recently developed parameter optimisation algorithm that is inspired by the foraging behaviour of honey bees. The Bees Algorithm performs a kind of exploitative neighbourhood search combined with random explorative search. Three neural networks were trained to reproduce a set of input/output numerical examples of the inverse kinematics of the main three joints of an articulated robotic manipulator. The results prove the remarkable robustness of the Bees Algorithm, which consistently trained the neural networks to model the kinematics data with very high accuracy. The learning results obtained by the proposed algorithm are compared to the results obtained by the standard Backpropagation Algorithm and an Evolutionary Algorithm. The comparative study highlights the superior performance of the proposed Bees Algorithm over the other algorithms.

In this paper, the Bee Algorithm is used to train a Neural Network. This is done by altering the connections and biases of the Neural Network (NN) so that the desired output from the input is obtained. The merging between the two concepts is tested to control an inverted pendulum which is a benchmark for testing control theories. The trained NN is used to stabilize the pendulum in its upright position. The NN is trained by comparing its response to that of a state feedback controller. The Bee Algorithm succeeded in training the NN for it to have the desired output. Moreover, the effect of changing the parameters of both the neural network and the bee algorithm is also studied.

Rolling element bearings are critical mechanical components in rotating machinery. Fault detection and diagnosis in the early stages of damage are necessary to prevent their malfunctioning and failure during operation. Vibration monitoring is the most widely used and cost-effective monitoring technique to detect, locate and distinguish faults in rolling element bearings. In this paper, an efficient method is proposed to extract optimizing features. The method employs capability features as well as the Bees Algorithm to obtain faults detection accurately and separably. This work presents an algorithm using optimum radial basis neural network by the use of the Bees Algorithm for automated diagnosis of localized faults in rolling element bearings. Optimum complementary capability values extracted from time-domain vibration signals are used as input features for the neural network. Optimum radial basis trained neural network are able to classify different states of the bearing with 100% accuracy. The proposed procedure requires only a few input features. Effectiveness of the proposed method is illustrated using the bearing vibration data obtained experimentally.

The current paper presents the use of the bees algorithm with Kalman filtering to train a radial basis function (RBF) neural network. An enhanced fuzzy selection system has been developed to choose local search sites depending on the error and training accuracy of the RBF network. The paper provides comparative results obtained when applying RBF neural classifiers trained using the new bees algorithm, the original bees algorithm, and the conventional RBF procedure to an industrial pattern classification problem.

Optimization of continuous synthesis of high purity carbon nanotubes (CNTs) using chemical vapour deposition (CVD) method was studied experimentally and theoretically. Iron pentacarbonyl (Fe(CO)5), acetylene (C2H2) and Ar were used as the catalyst source, carbon source and carrier gas respectively. The synthesis temperature and flow rates of Ar and acetylene were optimized to produce CNTs at a large scale. A flow rate of 30-120 sccm of acetylene and 500-3000 sccm of Ar at temperatures between 650-950 °C were examined. Using the fundamental trial and error method it was found that the maximum yield of pure CNTs can be produced at 750 °C with flow rates of 40-45 sccm of acetylene and 1500 sccm of Ar. In theoretical part, an artificial neural network (ANN) and the Bees Algorithm (BA) were used to model and optimize the CNTs production, based on the experimental data. The Bees Algorithm used the ANN as the fitness function and the optimum variables found as 60 sccm for acetylene, 555 sccm for argon and 759 °C for temperature. The computational results have relatively good agreement with the experimental results.

Recognition of unnatural patterns in process control charts through combining two types of neural networks

Available transfer capability is the most important factor for the identification of power transfer from gencos to discos. In a deregulated power system, power producer and customer share a common transmission network for wheeling of the electric power. This may cause violation of line flow, voltage and stability limits and thereby undermine the security limit [1]. To improve the available transfer capability of the system FACTS controllers are used. UPFC is the potential FACTS controller which is used to enhance ATC under deregulated environment. The optimal location for connecting FACTS controller and the amount of voltage and angle to be injected needs to be calculated. Here a hybrid technique is proposed to identify the optimal location for connecting UPFC and the amount of voltage and angle to be injected in the system. The neural network is used to identify the optimal location for connecting UPFC in the system and bees algorithm is used to identify the amount of voltage and angle to be injected in the system. The result shows the significant improvement in ATC and reducing total power losses in the system. Keywords: ATC (available transfer capability), UPFC (unified power flow controller), (BA) bees algorithm

ABSTRACTGenerating maximum possible power by wind turbines at low wind speeds is particularly important. Therefore, in this paper a hybrid method is presented for torque control in wind turbines. According to this method, the torque of wind turbine generator is set through a proportional and integral (PI) controller in such a way that at low wind speeds, the generated power by generator is maximized. In order to set PI controller gains to achieve maximum produced power by generator, multi-layer perceptron neural network is used. The set of optimized data is provided by Bees Algorithm in order to train this neural network. In order to evaluate the results of the proposed method, a 5 MW wind turbine made by NREL (National Renewable Energy Laboratory) is used. The results of simulations indicate the appropriate performance of the proposed method.

Heuristic-based neural networks for stochastic dynamic lot sizing problem

Unnatural patterns exhibited by control charts can be associated with certain assignable causes for process variation. Hence, accurately recognizing control chart patterns (CCPs) can significantly narrow down the scope of possible causes, and speeds up the troubleshooting process. This paper proposes a selective neural network (NN) ensemble approach DPSOEN, which employs a collection of several NNs trained for CCP identifications. DPSOEN provides more simple training and better performance than single NN. To further improve the performance of recognizers, several statistical features extracted from raw observations are used in the representation of input features. The simulation results indicate that integration of raw data and statistical features-based DPSOEN shows the best performance. Analysis from this study provides guidelines in developing NN ensemble-based SPC recognition systems.