Adaptive Neural Network Approach Research Articles

Adaptive deep homogenization theory for periodic heterogeneous materials

We present an adaptive physics-informed deep homogenization neural network (DHN) approach to formulate a full-field micromechanics model for elastic and thermoelastic periodic arrays with different microstructures. The unit cell solution is approximated by fully connected multilayers via minimizing a loss function formulated in terms of the sum of residuals from the stress equilibrium and heat conduction partial differential equations (PDEs), together with interfacial traction-free or adiabatic boundary conditions. In comparison, periodicity boundary conditions are directly satisfied by introducing a network layer with sinusoidal functions. Fully trainable weights are applied on all collocation points, which are simultaneously trained alongside the network weights. Hence, the network automatically assigns higher weights to the collocation points in the vicinity of the interface (particularly challenging regions of the unit cell solution) in the loss function. This compels the neural networks to enhance their performance at these specific points. The accuracy of adaptive DHN is verified against the finite element and the elasticity solution respectively for elliptical and circular cylindrical pores/fibers. The advantage of the adaptive DHN over the original DHN technique is justified by considering locally irregular porous architecture where pore–pore interaction makes training the network particularly slow and hard to optimize.

Practical fixed-time neural control for MIMO non-strict feedback nonlinear systems: an adaptive neural network approach

ABSTRACT This paper studies the non-singular practical fixed-time neural adaptive control issues for multi-input and multi-output (MIMO) nonlinear systems with non-strict feedback form. Neural networks (NN) are used to estimate the unknown nonlinearities and deal with the problem of an algebraic loop. Under the framework of the backstepping control design, a practical fixed-time adaptive NN control method is developed by using the adding power integration technology. According to the Lyapunov function theory, it is proved that the closed-loop system is practical fixed-time stable, and the system can track the desired reference signal within a fixed time. Finally, the proposed practical fixed-time control method is applied to a multi-motor control platform, which proves the effectiveness of the control method.

Novel hybrid optimization based adaptive deep convolution neural network approach for human activity recognition system

Novel hybrid optimization based adaptive deep convolution neural network approach for human activity recognition system

An analytical adaptive optimal control approach without solving HJB equation for nonlinear systems with input constraints

AbstractThis paper presents an analytical method to solve the optimal control problem for affine nonlinear systems with unknown drift dynamics. A new non‐quadratic cost function over an infinite horizon is presented that considers input constraints and includes the cost of the feed‐forward component of the control law. The mean value theorem for vector‐valued functions has been used to derive an integral form of this theorem. Based on this theorem, a rigorous proof is provided demonstrating that the cost function can be converted into another form. In the presence of input constraints, this converted form enables extracting the optimal control solution without solving the HJB equation. Additionally, unknown nonlinearity effects in drift dynamics are compensated in the control input. This is accomplished by estimating the unknown drift dynamics via an adaptive neural network (NN) approach. It is proven that the states and weights of NN are uniformly ultimately bounded based on a Lyapunov technique. The necessary and sufficient conditions are provided that ensure the optimality of the infinite horizon optimal control problem with a discount factor. As a result, it is demonstrated that the proposed approach satisfies the optimality criteria. To evaluate the effectiveness of the proposed approach, simulation examples are provided.

Adaptive Back-Propagation Neural Network and Particle Swarm Optimization-Based Approach for Optimizing the Output Power Biogas Fueled Electric Generator

Adaptive Back-Propagation Neural Network and Particle Swarm Optimization-Based Approach for Optimizing the Output Power Biogas Fueled Electric Generator

Vibration control for large space truss structure assembly using a distributed adaptive neural network approach

The vibration suppression problem of the large space truss structure during on-orbit assembly process is investigated in this paper, and a distributed adaptive neural network control approach is proposed to solve this problem. The dynamic modeling process of large space truss structure assembly is firstly developed. It can be used to describe the dynamic characteristics of the incrementally increasing space structure. To achieve the vibration suppression during on-orbit assembly, a distributed control approach is then proposed. The controller is designed for each assembly substructure, and a coordination term is added into the controller to reduce the interactions between the assembled substructures, and the distributed control system is therefore formed along with the assembly proceeds. The distributed controller is then redesigned using neural networks, to compensate the effect of the modeling uncertainties and the system parameter variation in the assembly process. It also could be noted that the dependence of controller design on dynamic models is then reduced. The simulation results demonstrate that the vibration is effectively suppressed, and the dynamic performance of the closed-loop system is improved. The controller designed based on the proposed dynamic model has good expansibility, and the control system also has robustness to external disturbance, and good fault tolerance when controller fails.

GPU-based similarity metrics computation and machine learning approaches for string similarity evaluation in large datasets

AbstractThe digital era brings up on one hand massive amounts of available data and on the other hand the need of parallel computing architectures for efficient data processing. String similarity evaluation is a processing task applied on large data volumes, commonly performed by various applications such as search engines, biomedical data analysis and even software tools for defending against viruses, spyware, or spam. String similarities are also used in musical industry for matching playlist records with repertory records composed of song titles, performer artists and producers names, aiming to assure copyright protection of mass-media broadcast materials. The present paper proposes a novel GPU-based approach for parallel implementation of the Jaro–Winkler string similarity metric computation, broadly used for matching strings over large datasets. The proposed implementation is applied in musical industry for matching playlist with over 100k records with a given repertory which includes a collection of over 1 million right owner records. The global GPU RAM memory is used to store multiple string lines representing repertory records, whereas single playlist string comparisons with the raw data are performed using the maximum number of available GPU threads and the stride operations. Further on, the accuracy of the Jaro–Winkler approach for the string matching procedure is increased using both an adaptive neural network approach guided by a novelty detection classifier (aNN) and a multiple-features neural network implementation (MF-NN). Thus, the aNN approach yielded an accuracy of 92% while the MF-NN approach achieved an accuracy of 99% at the cost of increased computational complexity. Timing considerations and the computational complexity are detailed for the proposed approaches compared with both the general-purpose processor (CPU) implementation and the state-of-the-art GPU approaches. A speed-up factor of 21.6 was obtained for the GPU-based Jaro–Winkler implementation compared with the CPU one, whereas a factor of 3.72 was obtained compared with the existing GPU implementation of string matching procedure based on Levenstein distance metrics.

Depth-Adaptive Deep Neural Network Based on Learning Layer Relevance Weights

In this paper, we propose two novel Adaptive Neural Network Approaches (ANNAs), which are intended to automatically learn the optimal network depth. In particular, the proposed class-independent and class-dependent ANNAs address two main challenges faced by typical deep learning paradigms. Namely, they overcome the problems of setting the optimal network depth and improving the model interpretability. Specifically, ANNA approaches simultaneously train the network model, learn the network depth in an unsupervised manner, and assign fuzzy relevance weights to each network layer to better decipher the model behavior. In addition, two novel cost functions were designed in order to optimize the layer fuzzy relevance weights along with the model hyper-parameters. The proposed ANNA approaches were assessed using standard benchmarking datasets and performance measures. The experiments proved their effectiveness compared to typical deep learning approaches, which rely on empirical tuning and scaling of the network depth. Moreover, the experimental findings demonstrated the ability of the proposed class-independent and class-dependent ANNAs to decrease the network complexity and build lightweight models for less overfitting risk and better generalization.

Adaptive Neural Network-Based Fixed-Time Tracking Controller for Disabilities Exoskeleton Wheelchair Robotic System

In this paper, an adaptive neural network approach is developed based on the integral nonsingular terminal sliding mode control method, with the aim of fixed-time position tracking control of a wheelchair upper-limb exoskeleton robot system under external disturbance. The dynamical equation of the upper-limb exoskeleton robot system is obtained using a free and typical model of the robotic manipulator. Afterward, the position tracking error between the actual and desired values of the upper-limb exoskeleton robot system is defined. Then, the integral nonsingular terminal sliding surface based on tracking error is proposed for fixed-time convergence of the tracking error. Furthermore, the adaptive neural network procedure is proposed to compensate for the external disturbance which exists in the upper-limb exoskeleton robotic system. Finally, to demonstrate the effectiveness of the proposed method, simulation results using MATLAB/Simulink are provided.

RBF Neural Network Sliding Mode Control for Passification of Nonlinear Time-Varying Delay Systems with Application to Offshore Cranes.

This paper is devoted to studying the passivity-based sliding mode control for nonlinear systems and its application to dock cranes through an adaptive neural network approach, where the system suffers from time-varying delay, external disturbance and unknown nonlinearity. First, relying on the generalized Lagrange formula, the mathematical model for the crane system is established. Second, by virtue of an integral-type sliding surface function and the equivalent control theory, a sliding mode dynamic system can be obtained with a satisfactory dynamic property. Third, based on the RBF neural network approach, an adaptive control law is designed to ensure the finite-time existence of sliding motion in the face of unknown nonlinearity. Fourth, feasible easy-checking linear matrix inequality conditions are developed to analyze passification performance of the resulting sliding motion. Finally, a simulation study is provided to confirm the validity of the proposed method.

Feedback Based Adaptive Recurrent Neural Network for Cancer Detection using Gene Data Pattern

Cancer detecting technology plays a vital role in the medical community. Researches have shown that patients that are affected by cancer carry same type of genetic patterns in their DNA. With this in mind, this research work concentrates on analysing gene pattern for detecting cancer using deep learning algorithms. The Feedback based Adaptive Recurrent Neural Network (FA-RNN) approach is designed to classify and analyse the gene pattern recognition. The data augmentation is done to improve the quality of the input data from COSMIC dataset which includes the detection of missing values, removing the noise present in input using multiple imputations and reducing higher base value can be done using dimensionality reduction process. After obtaining the improved dataset, the training phase begins by estimating the exact weight value of feedback layer using feedback weight loop calculation technique to lessen number of repetition during training. Moreover, the error calculation is done to evaluate the exact weight values of feedback layer used for classification. Finally the classification is done by selecting the next appropriate hidden neuron using the neuron selection activation function. The performance of the Feedback based Adaptive Recurrent Neural Network technique can be analysed using the evaluation metrics accuracy, computation time and Root Mean Square Error (RMSE) and the attained results are compared with the Recursive Neural Network(RNN) and Convolutional Neural Network(CNN) algorithms. The obtained results such as higher accuracy, reduced RMSE and less computation time in Feedback based Adaptive Recurrent Neural Network indicates that it performs the enhanced operation than CNN and RNN.

Adaptive Linear Neural Network Approach for Three-Phase Four-Wire Active Power Filtering under Non-Ideal Grid and Unbalanced Load Scenarios

This paper presents the enhancements performed on the adaptive linear neuron (ADALINE) technique so that it can be applied for active power filtering purposes in a three-phase four-wire system. In the context of active power filtering, the ADALINE technique which was initially developed for a single-phase two-wire system has been further expanded to suit three-phase three-wire system. For both systems, ADALINE techniques have been reported to be effective even when the grid voltage is distorted and/or unbalanced. However, further works that study the possibility to apply ADALINE technique in a three-phase four-wire system which invariably carries unbalanced loads, are rather limited. Hence, in this work, a control algorithm (named as enhanced-ADALINE) which combines the strength of highly selective filter (HSF), ADALINE concept and averaging function is proposed, to manage harmonics mitigation by shunt active power filter (SAPF) under non-ideal grid and unbalanced load scenarios. MATLAB-Simulink software is utilized to conduct an exhaustive simulation study which includes circuit connection of SAPF in a three-phase four-wire system, design of control algorithms, and performance assessments. Benchmarking with the existing algorithm is performed to examine the benefits of using the proposed algorithm. From the analysis, simulation findings are presented and thoroughly discussed to verify design concept, capability, and relevance of the proposed algorithm.

Improved Neural Network Control Approach for a Humanoid Arm

This study extended the knowledge over the improvement of the control performance for a seven degrees-of-freedom (7DOF) humanoid arm. An improved adaptive Gaussian radius basic function neural network (RBFNN) approach was proposed to ensure the reliability and stability of the humanoid arm control. Considering model uncertainties, the established dynamic model for the humanoid arm was divided into a nominal model and an error model. The error model was approximated by the RBFNN learning to compensate the uncertainties. The contribution of this study mainly concentrates on employing fruit fly optimization algorithm (FOA) to optimize the basic width parameter of the RBFNN, which can enhance the capability of the error approximation speed. Additionally, the output weights of the neural network were adjusted using the Lyapunov stability theory to improve the robustness of the RBFN-based error model. The simulation and experiment results demonstrate that the proposed approach is able to optimize the system state with less tracking errors, regulate the uncertain nonlinear dynamic characteristics, and effectively reduce unexpected interferences.

A New Spatial Algorithm Based on Adaptive Fuzzy Neural Network for Prediction of Crustal Motion Velocities in Earthquake Research

In earthquake studies, different methods are used in modeling of the crustal motions. In case of obscurity data structure, different approaches are needed in solving motion problems. In this paper, a new spatial algorithm has been developed which is based on adaptive fuzzy neural network (AFNN) approach for the prediction of the crustal motion velocities. In order to find the fuzzy class numbers regarding the network model formed by the fuzzification of the studied area, subtractive clustering algorithm is used. In determining the membership function, utilization of the variogram function which models the relationship that depends on distance among spatial data is proposed. The Marmara Region, Turkey, is used as the case for this study. In order to evaluate the performance of the approach, the kriging method is also utilized in the prediction and the results obtained from both methods are compared based on the mean-square-error criteria. It is observed that the AFNN approach yields results which are as effective as those of kriging. Consequently, it is shown that the AFNN approach will contribute to earthquake studies.

Stabilisation of a relative equilibrium of an underactuated AUV on SE(3)

ABSTRACTThis paper presents stabilisation of a relative equilibrium of an underactuated autonomous underwater vehicle (AUV) with three independent control inputs on Lie group SE(3), since SE(3) can describe the evolution of a rigid body without singularities or ambiguities. First, by using the interconnection and damping assignment (IDA) method through kinetic energy shaping, a controller is proposed to stabilise the relative equilibrium in the sense of Lyapunov stability in cotangent bundle T*SE(3) modulo an isotropy subgroup when a marine vehicle immerses in an ideal fluid. Then, for an AUV submersed in a viscous fluid, an asymptotical static control law consisting of three terms is designed to stabilise the relative equilibrium via the IDA method, which these three control terms are derived from kinetic energy shaping, potential energy shaping and damping reassignment, respectively. The control term derived from kinetic energy shaping preserves stability of this motion modulo translations along the shortest semi-axis in any direction and rotations about the same axis. Moreover, the static controller is developed to a dynamic one via adaptive neural network approach, which is robust for uncertainties in the system's damping and mass. Simulation results complete our work.

Prediction of damage parameters of a 3PL company via data mining and neural networks

This study covers an application in predicting error parameters for logistic service sector (logistic service provider). In the application real data regarding the last 5 years of a 3PL company has been used. The most significant criterion of the success of 3PL companies is the storage and transference of their products to their customers with no damage. It is for this reason that the data related to errors, in particular, have been dwelled upon. Data mining and ANN techniques used in practice are in widespread use. However, examples of their application in providing logistic service are few in number. For this purpose, the classification techniques of data mining were applied to estimate 3PL damage parameters as the class attribute. Find laws, decision tree and decision forest modules (engines) of Polyanalyst are used to discover the similarities and information recovery about the 3PL damage parameters. Moreover, in this study, to compare the prediction results artificial neural networks are used. In order to catch robust and adaptive neural network approach, four types of neural predictors are operated in this research. These neural network predictors are; Back Propagation Neural Network (BBNN), General Regression Neural Network, Radial Basis Neural Network and Adaptive Neuro-Fuzzy Neural Network (ANFIS). The results of four structures have proven that an ANFIS type can be employed to estimate the damage parameters of a 3PL company. This study shows how the 3PL can evaluate and dataof damages to improve their service quality and cost effectiveness to the customers. he analysis of the data related operational errors in 3PL’s is one of the most important contributions of this study in that such analyses can guide this and other similar companies in managing and reducing the number of their future errors. Analysis of these damages with artificial intelligence may help prevent 3PLs for the future process and also predict the effect of cost for the company.

Gas turbine modeling using adaptive fuzzy neural network approach based on measured data classification

The use of gas turbines is widespread in several industries such as; hydrocarbons, aerospace, power generation. However, despite to their many advantages, they are subject to multiple exploitation problem that need to be solved. Indeed, the purpose of the present paper is to develop mathematical models of this industrial system using an adaptive fuzzy neural network inference system. Where the knowledge variables in this complex system are determined from the real time input/output data measurements collected from the plant of the examined gas turbine. It is obvious that the advantage of the neuro-fuzzy modeling is to obtain robust model, which enable a decomposition of a complex system into a set of linear subsystems. On the other side, by focusing on the membership functions for residual generator to get consistent settings based on the used data structure classification and selection, where the main goal is to obtain a robust system information to ensure the supervision of the examined gas turbine.

An adaptive neural network approach to the tracking control of micro aerial vehicles in constrained space

ABSTRACTThis paper presents an adaptive neural network approach to the trajectory tracking control of micro aerial vehicles especially when they are flying in a limited indoor area. Differing from conventional controllers, the proposed controller employs the outer position loop to directly generate angular velocity commands in the presence of unknown aerodynamics and disturbances and then the fast inner loop to handle the angular rate control. Adaptive neural networks are deployed to estimate all the uncertain factors with the adaptation law derived from the Lyapunov function. To achieve a real-time performance, a norm estimation approach of ideal weights is designed to achieve a high bandwidth and lighten the burden of computation burden. Meanwhile, a barrier Lyapunov function is introduced to guarantee the constraint of vehicle positions as well as the validity of the neural network estimation. Simulations and practical flight tests are conducted to verify the feasibility and effectiveness of the proposed control strategy.

Comparison of gene regulatory networks using adaptive neural network and self-organising map approaches over Huh7 hepatoma cell microarray data matrix

Construction of gene regulatory network GRN is very important as it governs the expression levels of biomolecules in microarray data. In this article, we have developed GRNs by adaptive neural network ANN and self-organising map SOM approaches over Hepatitis C virus infection effect on Huh7 hepatoma cell microarray time series data. We then compared GRNs for the best performance analysis. We used fuzzy C-means clustering method to cluster the normalised dataset and then cluster centres are identified. After constructing GRNs within cluster centres, we analysed that SOM topology results a better performance providing minimum error to construct the GRN from sample data.

Comparison of gene regulatory networks using adaptive neural network and self-organising map approaches over Huh7 hepatoma cell microarray data matrix

Comparison of gene regulatory networks using adaptive neural network and self-organising map approaches over Huh7 hepatoma cell microarray data matrix