Implementation Of Artificial Neural Networks Research Articles

Neural networks within multi-core optic fibers

Hardware implementation of artificial neural networks facilitates real-time parallel processing of massive data sets. Optical neural networks offer low-volume 3D connectivity together with large bandwidth and minimal heat production in contrast to electronic implementation. Here, we present a conceptual design for in-fiber optical neural networks. Neurons and synapses are realized as individual silica cores in a multi-core fiber. Optical signals are transferred transversely between cores by means of optical coupling. Pump driven amplification in erbium-doped cores mimics synaptic interactions. We simulated three-layered feed-forward neural networks and explored their capabilities. Simulations suggest that networks can differentiate between given inputs depending on specific configurations of amplification; this implies classification and learning capabilities. Finally, we tested experimentally our basic neuronal elements using fibers, couplers, and amplifiers, and demonstrated that this configuration implements a neuron-like function. Therefore, devices similar to our proposed multi-core fiber could potentially serve as building blocks for future large-scale small-volume optical artificial neural networks.

Hardware Implementation of Artificial Neural Network for Data Ciphering

This paper introduces the design and realization of multiple blocks ciphering techniques on the FPGA (Field Programmable Gate Arrays). A back propagation neural networks have been built for substitution, permutation and XOR blocks ciphering using Neural Network Toolbox in MATLAB program. They are trained to encrypt the data, after obtaining the suitable weights, biases, activation function and layout. Afterward, they are described using VHDL and implemented using Xilinx Spartan-3E FPGA using two approaches: serial and parallel versions. The simulation results obtained with Xilinx ISE 9.2i software. The numerical precision is chosen carefully when implementing the Neural Network on FPGA. Obtained results from the hardware designs show accurate numeric values to cipher the data. As expected, the synthesis results indicate that the serial version requires less area resources than the parallel version. As, the data throughput in parallel version is higher than the serial version in rang between (1.13-1.5) times. Also, a slight difference can be observed in the maximum frequency.

Design and Implementation of High Speed Artificial Neural Network Based Sprott 94 S System on FPGA

FPGA-based embedding system designs have been preferred for industrial applications and prototyping because of the advantages of parallel processing, reconfigurability and low cost. Due to having characteristic structure of the parallel processing of Artificial Neural Networks (ANNs), these systems provide the advantage of speed and performance when they are implemented with FPGA-based hardware. The hardware implementation of transfer functions used for modeling non-linear systems is a challenging problem. Therefore, this problem creates convergence problems. In this paper, non-linear Sprott 94 S system has been modeled using ANNs running on FPGA. All related parameter values and processes are defined with IEEE-754-1985 32-bit floating point number format. ANN-based Sprott 94 S system design has been developed using VHDL synthesized using Xilinx ISE Design Tools. In test stage, ANN-based Sprott 94 S system has been tested using 3X100 data set and obtained error analysis results have been presented. The constructed design has been performed for Xilinx VIRTEX-6 family XC6VHX255T-3FF1923 FPGA chip using Place&Route process and chip usage statistics have been given. The clock frequency of ANN-based Sprott 94 S system which has pipeline processing scheme has been obtained with the value of 304.534 MHz. Accordingly, the proposed FPGA-based ANN system has produced 3X3.284 billion outputs in 1 second.

Condition monitoring of planetary gearbox by hardware implementation of artificial neural networks

The hazards of planetary gearboxes’ failures are the most crucial in the machinery which directly influence human safety like aircrafts. But also in an industry their damages can cause the large economic losses. Planetary gearboxes are used in wind turbines which operate in non-stationary conditions and are exposed to extreme events. Also bucket-wheel excavators are equipped with high-power gearboxes that are exposed to shocks. Continuous monitoring of their condition is crucial in view of early failures, and to ensure safety of exploitation. Artificial neural networks allow for a quick and effective association of the symptoms with the condition of the machine. Extensive research shows that neural networks can be successfully used to recognize gearboxes’ failures; they allow for detection of new failures which were not known at the time of training and can be applied for identification of failures in variable-speed applications. In a majority of the studies conducted so far neural networks were implemented in the software, but for dedicated engineering applications the hardware implementation is being used increasingly, due to high efficiency, flexibility and resistant to harsh environmental conditions. In this paper, a hardware implementation of an artificial neural network designed for condition monitoring of a planetary gearbox is presented. The implementation was done on a Field Programmable Gate Array (FPGA). It is characterized by much higher efficiency and stability than the software one. To assess condition of a gearbox working in non-stationary conditions and for chosen failure modes, a signal pre-processing algorithm based on filtration and estimation of statistics from the vibration signal was used. Additionally, the rewards-punishments training process was improved for a selected neural network, which is based on a Learning Vector Quantization (LVQ) algorithm. Presented classifier can be used as an independent diagnostic system or can be combined with traditional data acquisition systems using FPGAs.

Нейронные сети в системах управления преобразователями электрического подвижного состава

Objective: To improve the performance and accuracy of the system of voltage and frequency control for asynchronous traction motor (ATD) of rolling stock. Methods: To estimate the influence of artificial neural networks (INS) implementation to ATD control systems on the control indexes the comparison method was used. Results: The paper examines the possibility of INS implementation within the systems of direct control of traction electric motor moment. It also suggests a method of INS implementation as a ATD flow monitor. It provides the options of implementation and indicates the features of monitor structure design. Practical importance: Integration of the method of direct moment control and using of neural network technologies allows to improve the structure of converters and to improve the accuracy and the performance of output parameters, that, in its turn, decreases the power consumption, as well as losses in traction converters.

Hardware design and implementation of a novel ANN-based chaotic generator in FPGA

This paper presents a novel hardware implementation of Artificial Neural Networks (ANNs) for modeling of the Pehlivan–Uyaroglu Chaotic System (PUCS) on Field Programmable Gate Array (FPGA). There are two main parts in the proposed work. In the first part, a 3-8-3 Feed Forward Neural Network (FFNN) has been created using Matlab R2015a. The training results show that FFNN trained with back propagation algorithm exhibits satisfactory precision for the direct implementation. In the second part, the hardware implementation of the trained network has been carried out. The designed architecture is presented using Very High Speed Integrated Circuits Hardware Description Language (VHDL) and is implemented on a Xilinx Virtex 6 (XC6VCX240T) chip. All related parameters are defined with IEEE 754 single precision floating point number format. For the approximation of Log-Sigmoid transfer function, Xilinx's COordinate Rotation DIgital Computer (CORDIC) design has been employed. The design can be used with a clock frequency up to 266.429MHz. Finally, chip statistics of FPGA and analysis results have been presented. The proposed work have showed that chaotic systems can be successfully modeled using ANNs on FPGA. In future, chaos-based engineering applications can be performed using ANN-based chaotic oscillators on FPGA.

Reaction–diffusion chemistry implementation of associative memory neural network

Unconventional computing paradigms are typically very difficult to program. By implementing efficient parallel control architectures such as artificial neural networks, we show that it is possible to program unconventional paradigms with relative ease. The work presented implements correlation matrix memories (a form of artificial neural network based on associative memory) in reaction–diffusion chemistry, and shows that implementations of such artificial neural networks can be trained and act in a similar way to conventional implementations.

Improving quantitative structure-activity relationship models using Artificial Neural Networks trained with dropout.

Dropout is an Artificial Neural Network (ANN) training technique that has been shown to improve ANN performance across canonical machine learning (ML) datasets. Quantitative Structure Activity Relationship (QSAR) datasets used to relate chemical structure to biological activity in Ligand-Based Computer-Aided Drug Discovery pose unique challenges for ML techniques, such as heavily biased dataset composition, and relatively large number of descriptors relative to the number of actives. To test the hypothesis that dropout also improves QSAR ANNs, we conduct a benchmark on nine large QSAR datasets. Use of dropout improved both enrichment false positive rate and log-scaled area under the receiver-operating characteristic curve (logAUC) by 22-46% over conventional ANN implementations. Optimal dropout rates are found to be a function of the signal-to-noise ratio of the descriptor set, and relatively independent of the dataset. Dropout ANNs with 2D and 3D autocorrelation descriptors outperform conventional ANNs as well as optimized fingerprint similarity search methods.

Software Design Challenges in Time Series Prediction Systems Using Parallel Implementation of Artificial Neural Networks.

Software development life cycle has been characterized by destructive disconnects between activities like planning, analysis, design, and programming. Particularly software developed with prediction based results is always a big challenge for designers. Time series data forecasting like currency exchange, stock prices, and weather report are some of the areas where an extensive research is going on for the last three decades. In the initial days, the problems with financial analysis and prediction were solved by statistical models and methods. For the last two decades, a large number of Artificial Neural Networks based learning models have been proposed to solve the problems of financial data and get accurate results in prediction of the future trends and prices. This paper addressed some architectural design related issues for performance improvement through vectorising the strengths of multivariate econometric time series models and Artificial Neural Networks. It provides an adaptive approach for predicting exchange rates and it can be called hybrid methodology for predicting exchange rates. This framework is tested for finding the accuracy and performance of parallel algorithms used.

Investigating camera calibration for eye tracking of the physically challenged

This paper tries to address a problem faced by a proportion of community - the paralysed people, and investigates a solution in calibrating a camera used for the purpose of eye detection. The eye tracking is later used to read characters from a partner-assisted keyboard. As such, this keyboard solves the problem of immobility. In order to calibrate camera for the purpose of eye detection, two approaches are implemented-Artificial Neural Networks and Support Vector Machines in context of paralysed people, and are discussed in detail in this paper. The results of both approaches are compared in order to select the best approach to be used in keyboard that operates on eye tracking.

Implementation of artificial neural-networks to model the performance parameters of Stirling engine

The Stirling engine is defined as a simple form of external-combustion engine which employs a compressible working fluid. From a theoretical point of view the Stirling engine can be very effective at Carnot efficiency to convert heat into mechanical work. It is an environmental friendly heat engine which could reduce CO2 emission through combustion process. Performance of Stirling engine changes via changing several variables and the aforementioned variables should be optimized to accomplish maximum performance of Stirling engine. Among all the variables representing the performance of Stirling engine, output power and shaft torque are most common for illustrating the performance of Stirling engines. In this research predicting two aforementioned variables with high accuracy and flexibility is investigated. To gain this goal, a predictive and easy-to-use tool is developed based on the concept of the artificial neural network (ANN) to predict output power and shaft torque of the Stirling engine. Furthermore, precise data samples from previous researches are employed to construct this easy-to-use model. Based on the outputs obtained from an easy-to-use model developed in this research, ANN model could help experts in designing of Stirling engine with low degree of uncertainty.

FPGA-Based Stochastic Echo State Networks for Time-Series Forecasting.

Hardware implementation of artificial neural networks (ANNs) allows exploiting the inherent parallelism of these systems. Nevertheless, they require a large amount of resources in terms of area and power dissipation. Recently, Reservoir Computing (RC) has arisen as a strategic technique to design recurrent neural networks (RNNs) with simple learning capabilities. In this work, we show a new approach to implement RC systems with digital gates. The proposed method is based on the use of probabilistic computing concepts to reduce the hardware required to implement different arithmetic operations. The result is the development of a highly functional system with low hardware resources. The presented methodology is applied to chaotic time-series forecasting.

Neural Network Reinforcement Learning for Walking Control of a 3-Link Biped Robot

In recent years, researches on adaptive control have focused on bio-inspired learning techniques to deal with real-life applications. Reinforcement Learning (RL) is one of these major techniques, which has been widely used in robot control approaches. The implementation of artificial neural networks in RL algorithms enables more efficient optimal control strategies. This article proposes a field application of neural network reinforcement learning (NNRL) for walking control of an active simulated 3-link biped robot. The adaptive control agent consists of two neural network units, known as actor and critic for learning prediction and learning control tasks. Results of the presented control method reveal its efficiency in stable walking control of the biped robot model as a nonlinear complex dynamic task.

Predicting the moment capacity of RC beams exposed to fire using ANNs

This research investigates the implementation of artificial neural networks (ANNs) to estimate the moment capacity (Mr) of reinforced concrete (RC) beams under rising temperatures due to fire. 280 data were obtained for ANN model. Input layer in ANN model consisted of eight input parameters; the beam width (bw), the beam depth (d), the ratio of (bw/d), distance from the beam edge to the center of the rebar (d′), the ratio of (d′/d), fire time (texposure), the reinforcement area (Ast), and concrete compressive strength (fc). It is shown that the ANN model can be used to predict the Mr of RC beams exposed to fire with high accuracy. The predicted Mr by ANN are consistent with the results obtained using Mr equation. It was observed from the results the ANN model reduces the computational complexity problem in determining Mr. Consequently, the ANN model was used to examine the effects of the inputs parameters on Mr.

Automatic generation of LUTs for Hardware Neural Networks

One very important component of Hardware implementations of Artificial Neural Networks (HNN) is the activation function. A common solution for activation function design in HNN is the Look-Up Table (LUT) representation that has acceptable area constraints and usually achieves acceptable execution time. Tools and techniques for automatic generation of LUTs have been developed but have some problems as high complexity, high dependency of external tools or languages and the optimization techniques that are not efficient enough for critical applications. In order to solve some of the presented problems this paper describes a method for automatic LUT generation that overcomes some deficiencies of existing methods using a simple approach. This method was evaluated initially considering the xor function and the sine function for evaluation purposes. After these initial tests, two other ANN based robotic systems were tested. These robotic systems will be implemented in hardware in the near future and they demonstrate a different behavior using different domains of the same activation function. Proposed method achieved good results that are directly related to how neural networks define the activation function domain.

A preliminary study on the role of acoustic emission on inferring Cerchar abrasivity index of rocks using artificial neural network

The wear rate of working tools during the cutting and drilling of rocks is closely related to the abrasiveness of those rocks. As the contact area of the tools increases, due to wear, the specific cutting energy will also increase, and that directly affects the overall consumption of excavation tools. A new artificial intelligence (AI) based model has been developed. It utilizes acoustic emission (AE) and rock properties as main indicators of rock abrasivity, estimated by Cerchar Abrasivity Index (CAI). AE sensors are attached to both the Cerchar testing apparatus and the rock in question while conducting scratch tests using hardened steel pins of 42 and 56 HRC. Prior to the implementation of Artificial Neural Network (ANN) modeling, the selection of independent variables was carried out via Gamma test and V-ratio analyses. As a result, AE parameters, such as total number of events and root mean square of signal, in addition to testing parameters (i.e. uniaxial compressive strength, Young׳s Modulus, quartz content and pin hardness) are found to be the optimum model input combination needed to accurately predict CAI.

Digital Artificial Neural Network Implementation on a FPGA for data classification

This paper presents the implementation of a Digital Artificial Neural Network (ANN) of type Multilayer Perceptron trained for data classification corresponding to the output characteristic of a transistor MOSFET. The ANN digital was build and training with an adequate number of neurons and samples for achieve an approximate of 90% correct data classification of the transistor MOSFET.

MCDM Tools Application for Selection of Suppliers in Manufacturing Industries Using AHP, Fuzzy Logic and ANN

In manufacturing industry selection of right supplier (s) is one of the challenging activities. Supplier selection is the method of determining the appropriate suppliers to provide the right quality of products, at the right quantities, at the right price in the right time. During the last ten years based on the changes in the business sector, Multi-Criteria Decision Making (MCDM) has been one of the fastest growing areas. By choosing the best supplier, the companies can increase their productivity and also competitive advantage. In the earlier period, several methods have been proposed to solve the supplier selection problem, such as the linear weighting methods, the analytic network process, total cost approaches and mathematical programming techniques. Although the linear programming methods are simple, they are not accurate and more variations exist. So the latest multi-criteria decision making tools such as Analytical Hierarchy Process (AHP), Fuzzy Logic (FL) and Artificial Neural Network (ANN) are deeply studied. While selecting the best supplier, many criteria are involved. The criteria may be different for different approach. The decision makers always express their preferences on alternatives or on the attributes of supplier which can be used to help rank the supplier and choose the enviable one. In this article, application of fuzzy logic and AHP are discussed with case-studies and also presented the discussion about implementation of ANN for supplier selection.

Modular implementation of artificial neural network in predicting in-flight particle characteristics of an atmospheric plasma spray process

This paper presents a modular implementation of an artificial neural network to model the atmospheric plasma spray process in predicting the in-flight particle characteristics from the input processing parameters. The in-flight particle characteristics influence the structure and properties of the thermal spray coating and, thus, are considered important parameters to comprehend, simulate and predict the manufacturing process. The modular implementation allows simplification of the optimized model structure with enhanced ability to generalise the network. As well, the underlying relationship between each of the output in-flight characteristics with respect to the input processing parameters is explored. Smaller networks are constructed that achieves better, or in some cases, similar results. The training process is found to be more robust and stable along with fewer fluctuations in the values of the network parameters. The networks also respond to the variations of the number of hidden layer neurons with some definite trend. The predictable trend enhances reliability of the application of the artificial neural network in modelling the atmospheric plasma spray process and overcomes the variability and non-linearity associated with the process.

Time-order error and scalar variance in a computational model of human timing: simulations and predictions

This work introduces a computational model of human temporal discrimination mechanism – the Clock-Counter Timing Network. It is an artificial neural network implementation of a timing mechanism based on the informational architecture of the popular Scalar Timing Model. The model has been simulated in a virtual environment enabling computational experiments which imitate a temporal discrimination task – the two-alternative forced choice task. The influence of key parameters of the model (including the internal pacemaker speed and the variability of memory translation) on the network accuracy and the time-order error phenomenon has been evaluated. The results of simulations reveal how activities of different modules contribute to the overall performance of the model. While the number of significant effects is quite large, the article focuses on the relevant observations concerning the influence of the pacemaker speed and the scalar source of variance on the measured indicators of network performance. The results of performed experiments demonstrate consequences of the fundamental assumptions of the clock-counter model for the results in a temporal discrimination task. The results can be compared and verified in empirical experiments with human participants, especially when the modes of activity of the internal timing mechanism are changed because of some external conditions, or are impaired due to some kind of a neural degradation process.