Recursive Orthogonal Least Square Based Soft Sensor for Batch Distillation

Novel constructive and destructive parsimonious extreme learning machines (CP- and DP-ELM) are proposed in this paper. By virtue of the proposed ELMs, parsimonious structure and excellent generalization of multiinput-multioutput single hidden-layer feedforward networks (SLFNs) are obtained. The proposed ELMs are developed by innovative decomposition of the recursive orthogonal least squares procedure into sequential partial orthogonalization (SPO). The salient features of the proposed approaches are as follows: 1) Initial hidden nodes are randomly generated by the ELM methodology and recursively orthogonalized into an upper triangular matrix with dramatic reduction in matrix size; 2) the constructive SPO in the CP-ELM focuses on the partial matrix with the subcolumn of the selected regressor including nonzeros as the first column while the destructive SPO in the DP-ELM operates on the partial matrix including elements determined by the removed regressor; 3) termination criteria for CP- and DP-ELM are simplified by the additional residual error reduction method; and 4) the output weights of the SLFN need not be solved in the model selection procedure and is derived from the final upper triangular equation by backward substitution. Both single- and multi-output real-world regression data sets are used to verify the effectiveness and superiority of the CP- and DP-ELM in terms of parsimonious architecture and generalization accuracy. Innovative applications to nonlinear time-series modeling demonstrate superior identification results.

A localised forgetting method for on-line adaptation of gaussion RBFN models

Distillation anomaly and fault detection based on unsupervised dimensionality principal component analysis method

In this paper, we propose a novel complex-valued RBF networks approach for dynamic behavioral modeling of nonlinear power amplifier with memory. The complex recursive orthogonal least squares (ROLS) algorithm is applied to train complex radial basis function (RBF) network when modeling, so as to save large memory and computational efforts. Using the information available from the trained network with complex ROLS algorithm, the effective centers of the network can be obtained by adopting backward selection algorithm, which achieve acceptable accuracy with significant reduction of network structure. The model has been validated using measured and simulated data. The results of simulation and experimentation show that this complex RBF network model requires a significantly reduced complexity in the analysis and training procedures, when driven with WCDMA signals, than previously published neural-network-based PA models.

This paper presents a novel power quality disturbance detection and classification method of distribution power system based on complex wavelet transform (WT) and radial basis function (RBF) neural network. The complex supported orthogonal wavelets is employed to extract the feature information of disturbance signal, and finally proposed to explore several novel wavelet combined information (CI) to analyze the disturbance, superior to real wavelet analysis result. The feature obtained from WT coefficients are inputted into RBF network for power quality disturbance pattern classification. The power quality disturbance classification model is established and the synthesized method of recursive orthogonal least squares algorithm (ROLSA) with improved givens transform is used to fulfill the network structure parameters. By means of choosing enough samples to train the recognition model, the type of disturbance can be obtained when signal representing fault is inputted to the trained network. The simulation results demonstrate that the complex WT combined with RBF network are more sensitive to signal singularity, and found to be significant improvement for acquiring signal feature information.

Recursive orthogonal least squares (ROLS) is a numerically robust method for solving for the output layer weights of a radial basis function (RBF) network, and requires less computer memory than the batch alternative. In this paper, the use of ROLS is extended to selecting the centers of an RBF network. It is shown that the information available in an ROLS algorithm after network training can be used to sequentially select centers to minimize the network output error. This provides efficient methods for network reduction to achieve smaller architectures with acceptable accuracy and without retraining. Two selection methods are developed, forward and backward. The methods are illustrated in applications of RBF networks to modeling a nonlinear time series and a real multiinput-multioutput chemical process. The final network models obtained achieve acceptable accuracy with significant reductions in the number of required centers.

Chapter 11 - Distillation

Chapter 6 - Modeling and optimizing dynamic networks: Applications in process engineering and energy supply

A linear parameter-varying model (LPVM) is developed for nonlinear dynamic systems using a radial basis function (RBF) neural network. The training of the LPVM is formulated as the least-squares problem and the recursive orthogonal least-squares algorithm is applied. Model adaptation is also developed with a localized forgetting method for on-line weight updating. The LPVM-based model predictive control (MPC) is developed and the convexity of the optimization is preserved. The developed LPVM is applied to a laboratory-scaled chemical reactor rig. The real data modelling and on-line control implementation are presented. Decentralized proportional–integral–derivative (PID) control is also designed and implemented for the reactor for comparison. The superiority of the tracking performance by the LPVM-based MPC over that by PID control is clearly demonstrated. The performance using the fixed LPVM is also improved by using the adaptive model.

Batch distillation for separating the acetone and n-heptane binary azeotrope mixture: Optimization and simulation

Inferential model-based control of multicomponent batch distillation

Two contributions on 3D implicit surface reconstruction from scattered points are presented in this paper. Firstly, least square radial basis functions (LS RBF) are deduced from the conventional RBF formulations, which makes it possible to use fewer centers when reconstruction. Then we use orthogonal least square (OLS) method to select significant centers from large and dense point data sets. From the selected centers, an implicit continuous function is constructed efficiently. This scheme can overcome the problem of numerical ill-conditioning of coefficient matrix and over-fitting. Experimental results show that our two methods are efficient and highly satisfactory in perception and quantification.

Radial Basis Function Network Configuration Using Mutual Information and the Orthogonal Least Squares Algorithm

<p>An investigation into optimized support vector machine (SVM) integrated with principal component analysis (PCA) and orthogonal least square (OLS) in classifying asphyxiated infant cry was performed in this study. Three approaches were used in the classification; SVM, PCA-SVM, and OLS-SVM. Various numbers of features extracted from Mel-frequency Cepstral coefficient (MFCC) were tested to obtain the optimal parameters of SVM kernels. Once the optimal feature set is obtained, PCA and OLS selected the most significant features and the optimized SVM then classified the selected cry patterns. In PCA-SVM, eigenvalue-one-criterion (EOC), cumulative percentage variance (CPV) and the Scree test (SCREE) were used to select the most significant features. SVM with radial basis function (RBF) kernel was chosen in the classification stage. The classification accuracy and computation time were computed to evaluate the performance of each method. The best method for classifying asphyxiated infant cry is PCA-SVM with EOC since it produces the highest classification accuracy which is 94.84%. Using PCA-SVM, the classification process was performed in 1.98s only. The results also show that employing feature selection techniques could enhance the classifier performance.</p>

\nThe availability of liquid waste of rise flour can be maintained in continuously and abundant, and is one of waste water that is underutilized, but liquid waste of rise flour has a starch content which can be used as one of the producers of ethanol. This study has the objective to produce bioethanol and a prototype bio-ethanol industry, and perform simulations before conducting experiments, as well as a reference in developing the bio-ethanol industry in Indonesia.\nBased on preliminary analysis obtained starch content of the rice is about 79% b/b, while the starch content of the liquid waste of rise flour is quite high about 30% v/v. The process of preliminary research that has been done is the hydrolysis with Bacillus then followed by bacterial fermentation of glucose by Saccharomyces cerevisiae. The quality of ethanol produced is influenced by the levels of starch, glucose, acidity (pH), the volume ratio of waste with Bacillus volume, the number of bacteria Saccharomyces cerevisiae, fermentation time that has been done in the preliminary study. The filtrate resulting from the hydrolysis and fermentation processes of contains ethanol about 16% v/v carried out a batch distillation process to obtain bioethanol, variables in a batch distillation process is ethanol composition and temperature. Before experimental batch distillation process simulation ternary system Ethanol-Water-Impurities (EWI) with a model of Differential-Algebraic-Equations (DAEs), using the programming language Mathlab version 7 in advance, so that in the determination of the variables can be more effective and more research costs cheap.\nThe results obtained experimentally study the composition of bioethanol a maximum of 95% to 96%, the results of experiments and simulations EWI ternary system form the temperature profile, the profile of the composition of liquid and vapor composition profile dimensionless time functions both at the bottom and in the distillate shows the results of the same approach. The simulation results before used reference in experiments performed the validation beforehand, so that the ternary system simulation EWI after validation of reference can be used in experiments.\n