Optimal Neural Network Topology Research Articles

Convolution neural network application for first‐break picking for land seismic data

ABSTRACTAn automatic and robust algorithm for the first‐break picking is necessary to build the near‐surface velocity model. We propose the algorithm based on a convolution neural network. The introduced first‐break picking strategy and neural network architecture are suited for processing large volumes of seismic exploration data with reasonable computational resources. To develop an optimal neural network topology and architecture, extensive testing was performed. We compared several architectures of neural networks, including one‐ and two‐dimensional approaches. Our tests justify that the one‐dimensional approach (trace‐by‐trace processing) provides the most reliable results in the case of first‐break travel‐time variations typical of complicated near‐surface structures. This study demonstrates that the four‐layered neural network trained on 5,000 traces is enough for robust first‐break picking. The algorithm is evaluated on two land‐acquisition field datasets from West Siberia with a total used size of about 7 million traces. The first dataset is used for training, and the second one is used only for testing. For both datasets, the error between the original and the predicted first breaks is not more than three samples for 95% of traces. The final evaluation is done by a comparison of seismic stacks to prove the benefits of the approach and its robustness for offsets over 600 m. Finally, the influence of choosing the locations for the training dataset is discussed, and a strategy for using the proposed approach in production work is introduced.

Deep learning based phase transformation model for the prediction of microstructure and mechanical properties of hot-stamped parts

Hot stamping of boron steel is an effective way for weight reduction of automobiles parts, but suffers from the loss of mechanical properties due to incomplete martensitic transformation. Different from general heat treatment, phase transformation in hot stamping relates to both thermal loading path and deformation history. Unfortunately, the existing models have difficulty in capturing the history-dependent phase transformation behavior of boron steel under hot stamping conditions. In the present work, the gated recurrent unit (GRU) are combined with fully connected neural network (FCNN) to capture the coupling effect of both nonlinear strain history and thermal loading path on phase transformation. A unified deep-learning based phase transformation (DLPT) model is developed to integrate both diffusive and diffusionless transformation to cover all feasible thermo-mechanical loading path in hot stamping process. A hybrid driven thermo-mechanical-metallurgical (TMM) framework is developed by integrating the DLPT model into general TMM framework, which providing a novel approach for two-scale simulation of hot stamping process. A comprehensive database of phase transformation is constructed for model training based on well-designed thermo-mechanical loading experiments to cover hot stamping conditions, and the optimal topology of neural network in DLPT model is obtained by training on this database. The accuracy and reliability of DLPT model is systematically demonstrated by dieless hot V-bending and hot stamping of T-shaped parts besides CCT and DCCT tests. The net effect of local plastic strain on subsequent phase transformation is clarified with accuracy by dieless V-bending with the removal of disturbing factors. Non-uniform distribution in hardness of the final hot stamped parts can be predicted with the DLPT model successfully, which is better in accuracy than in case of working with the conventional model.

P-Wave Sonic Log Predictive Modeling with Optimal Artificial Neural Networks Topology

Given the financial challenges facing the oil and gas industry, the value of the information is considered relatively high; therefore, data science has been an alternative compensating tool. This study aimed to find an optimal neural network topology that provides an ideal data solution by studying neural network topology. Therefore, we trained different neural networks topologies in terms of the number of hidden neurons and layers. Volve oil field data is used in this study to predict the compressional sonic wave travel time. Optimal Neural Network topology found using five hidden layers and five hidden neurons while using a single layer with different numbers of hidden neurons was ineffective. The highest training and testing accuracy with a single hidden layer found 0.94 and 0.914, respectively. In contrast, it was found 0.947 and 0.934 with 50 hidden neurons and five hidden layers. Yet, increasing the number of hidden layers and hidden neurons is found to cause overfitting; therefore, only an optimal topology is a critical factor.

Neural networks for predicting the temperature-dependent viscoelastic response of PEEK under constant stress rate loading

High-performance polymer composites are used in demanding applications in civil and aerospace engineering. Often, structures made from such composites are monitored using structural health monitoring systems. This investigation aims to use a multilayer perceptron neural network to model polymer response to a non-standard excitation under different temperature conditions. Model could be implemented into health monitoring systems. Specifically, the neural network was used to model PEEK material's creep behavior under constant shear stress rate excitation at different temperatures. Optimal neural network topology, the effect of the amount of training data and its distribution in a temperature range on prediction quality were investigated.The results showed that based on the proposed optimization criterion, a properly trained neural network can predict polymeric material behavior within the experimental error. The neural network also enabled good prediction at temperatures where stress-strain behavior was not experimentally determined.

Multiobjective PSO based adaption of neural network topology for pixel classification in satellite imagery

The proposed work involves the multiobjective PSO based adaption of optimal neural network topology for the classification of multispectral satellite images. It is per pixel supervised classification using spectral bands (original feature space). This paper also presents a thorough experimental analysis to investigate the behavior of neural network classifier for given problem. Based on 1050 number of experiments, we conclude that following two critical issues needs to be addressed: (1) selection of most discriminative spectral bands and (2) determination of optimal number of nodes in hidden layer. We propose new methodology based on multiobjective particle swarm optimization (MOPSO) technique to determine discriminative spectral bands and the number of hidden layer node simultaneously. The accuracy with neural network structure thus obtained is compared with that of traditional classifiers like MLC and Euclidean classifier. The performance of proposed classifier is evaluated quantitatively using Xie-Beni and β indexes. The result shows the superiority of the proposed method to the conventional one.

FINDING PROBLEM SPECIFIC SHANNON INFORMATION IN HIGH DIMENSIONAL INPUT SPACE FOR ARTIFICIAL NEURAL NETWORKS

Artificial neural networks, due to their ability to find the underlying model even in complex highly nonlinear and highly coupled problems, have found significant use as prediction engines in many domains. However, in problems where the input space is of high dimensionality, there is the unsolved problem of reducing dimensionality in some optimal way such that Shannon information important to the prediction is preserved. The important Shannon information may be a subset of total information with an unknown partition, unknown coupling and linear or nonlinear in nature. Solving this problem is an important step in classes of machine learning problems and many data mining applications. This paper describes a semi-automatic algorithm that was developed over a 5-year period while solving problems with increasing dimensionality and difficulty in (a) flow prediction for a magnetically levitated artificial heart (13 dimensions), (b) simultaneous chemical identification/concentration in gas chromatography (22 detection dimensions with wavelet compressed time series of 180,000 points), and finally in (c) financial analytics portfolio prediction in credit card and sub-prime debt problems (80 to 300 dimensions of sparse data with a portfolio value of approximately US$300,000,000.00). The algorithm develops a map of input space combinations and their importance to the prediction. This information is used directly to construct the optimal neural network topology for a given error performance. Importantly, the algorithm also produces information that shows whether the space between input nodes is linear or nonlinear; an important parameter in determining the number of training points required in the reduced dimensionality of the training set. Software was developed in the MatLAB environment using the Artificial Neural Network Toolbox, Parallel and Distributed Computing toolboxes, and runs on Windows or Linux based supercomputers. Trained neural networks can be compiled and linked to server applications and run on normal servers or clusters for transaction or web based processing. In this paper, application of the algorithm to two separate financial analytics prediction problems with large dimensionality and sparse data sets are shown. The algorithm is an important development in machine learning for an important class of problems in prediction, clustering, image analysis, and data mining. In the first example application for subprime debt portfolio analysis, performance of the neural network provided a 98.4% prediction rate, compared to 33% rate using traditional linear methods. In the second example application regarding credit card debt, performance of the algorithm provided a 95% accurate prediction (in terms of match rate), and is 10% better than other methods we have compared against, primarily logistic regression.

Neural Network Control Chart Architecture for Monitoring Non-Conformities in a Poisson Process

The uses of Neural Network (NN) models have recently been recommended as statistical quality control (SQC) tools. The advantages of NNs, particularly the robustness of the nonlinear modeling abilities, are appealing to quality control practitioners for use in process monitoring. Advances in computing power have also made the Neural Network Control Charts (NNCC) an alternative SQC technique.The systematic Design of Experiment (DOE) methodology is employed to find near optimal NN topology for NNCC for Poisson data. A (2k) full factorial design is implemented and supplemented as needed to investigate NN topologies. The effect of the following factors were investigated through a simulation study: the number of the inputs “n”, the number of nodes in the hidden layer(s), the training data size, and in-control mean for shift range 0-3σ . The guidelines and steps of constructing the DOE study for the NNCC is given, along with an example.

Optimization methodology based on neural networks and self-adaptive differential evolution algorithm applied to an aerobic fermentation process

The determination of the optimal neural network topology is an important aspect when using neural models. Due to the lack of consistent rules, this is a difficult problem, which is solved in this paper using an evolutionary algorithm namely Differential Evolution. An improved, simple, and flexible self-adaptive variant of Differential Evolution algorithm is proposed and tested. The algorithm included two initialization strategies (normal distribution and normal distribution combined with the opposition based principle) and a modified mutation principle. Because the methodology contains new elements, a specific name has been assigned, SADE-NN-1. In order to determine the most influential inputs of the models, a sensitivity analysis was applied. The case study considered in this work refer to the oxygen mass transfer coefficient in stirred bioreactors in the presence of n-dodecane as oxygen vector. The oxygen transfer in the fermentation broths has a significant influence on the growth of cultivated microorganism, the accurate modeling of this process being an important problem that has to be solved in order to optimize the aerobic fermentation process.The neural networks predicted the mass transfer coefficients with high accuracy, which indicates that the proposed methodology had a good performance. The same methodology, with a few modifications, and with the best neural network models, was used for determining the optimal conditions for which the mass transfer coefficient is maximized.A short review of the differential evolution methodology is realized in the first part of this article, presenting the main characteristics and variants, with advantages and disadvantages, and fitting in the modifications proposed within the existing directions of research.

Use of genetic algorithm to design optimal neural network structure

In this research, neural network (NN) and genetic algorithm (GA) are used together to design optimal NN structure. The proposed approach combines the characteristics of GA and NN to reduce the computational complexity of artificial intelligence applications in design and manufacturing. Genetic input selection approach is introduced to obtain optimal NN topology. Experimental results are given to evaluate the performance of the proposed system.

A genetic algorithm to obtain the optimal recurrent neural network

Selecting the optimal topology of a neural network for a particular application is a difficult task. In the case of recurrent neural networks, most methods only induce topologies in which their neurons are fully connected. In this paper, we present a genetic algorithm capable of obtaining not only the optimal topology of a recurrent neural network but also the least number of connections necessary. Finally, this genetic algorithm is applied to a problem of grammatical inference using neural networks, with very good results.