Influence Of Network Parameters Research Articles

Semantic segmentation of large-scale segmental lining point clouds using 3D deep learning

It is highly challenging to comprehensively measure the deformations of large-scale tunnel structures because of the numerous individual segments of tunnel linings. Laser scanning and 3D deep learning (DL) can provide abundant geometric data and the ability to interpret such data automatically. To fill the gap in applying 3D DL to the segment-wise segmentation of large-scale tunnel point clouds, a new 3D dataset and an optimized DL network are presented in this paper. The point cloud dataset was manually annotated by labeling every individual segment and element. Subsequently, a 3D DL network capable of handling large-scale point clouds (approximately 106 points) was developed to perform the data experiments. The influences of network parameters, namely, the number of neighbor points, layer number, down-sampling ratio, and channel number, were investigated in the data experiments. The results indicate that the number of neighboring points and channel number have a notable influence on the segmentation performance, with the best mean interaction over union reaching 85.5%. This study provides preliminary insights into the 3D DL network design for large-scale tunnel point clouds.

Influence of network parameters on the accuracy of control and diagnostics of resistance welding

The influence of parasitic parameters of the power supply network (active resistance and inductance) on the accuracy of resistance welding diagnostics is estimated. Evaluation of the effectiveness of various control algorithms when operating under conditions of fluctuations in the supply voltage is performed. The methodic for determining the parasitic parameters of the network is presented. Keywords: resistance welding, automation, control and diagnostics. KlimovTGU@yandex.ru

Combined formаtion of a cryptographic key using synchronized artificial neural networks

А combined method for forming a cryptographic key is proposed in the article. The proposed combined formation consists of two stages: the formation of partially coinciding binary sequences using synchronized artificial neural networks and the elimination of mismatched bits by open comparison of the parities of bit pairs. In this paper, possible vulnerabilities of the basic method of forming a cryptographic key using synchronized artificial neural networks are considered, their danger is assessed, and a correction of the method is proposed to ensure the required confidentiality of the generated shared secret. At the first stage, a deferred brute-force attack is considered. To neutralize this attack, it is proposed to use the convolution function of the results of several independent synchronizations. As a convolution function, the bitwise addition modulo 2 of the vectors of the weights of the networks is used. Due to the correction of the first stage of the basic algorithm, the amount of deferred search exponentially increases, and frequency analysis of binary sequences also becomes ineffective. At the second stage, an attack based on the knowledge of pair parities is considered, taking into account the proposed method for correcting the first stage. The analysis of the influence of network parameters on the process of eliminating the bit mismatch at the second stage is carried out. Statistical modeling of this analysis has been performed. The results obtained showed that the cryptanalyst could not uniquely distinguish the values of the remaining bits. The proposed combined method makes it possible to increase the confidentiality of the generated shared secret and significantly reduce the number of information exchanges in comparison with the Neural key generation technology.

Charge and Thermoelectric Transport in Polymer-Sorted Semiconducting Single-Walled Carbon Nanotube Networks.

Understanding the charge transport mechanisms in chirality-selected single-walled carbon nanotube (SWCNT) networks and the influence of network parameters is essential for further advances of their optoelectronic and thermoelectric applications. Here, we report on charge density and temperature-dependent field-effect mobility and on-chip field-effect-modulated Seebeck coefficient measurements of polymer-sorted monochiral small-diameter (6,5) (0.76 nm) and mixed large-diameter SWCNT (1.17-1.55 nm) networks (plasma torch nanotubes, RN) with different network densities and length distributions. All untreated networks display balanced ambipolar transport and electron-hole symmetric Seebeck coefficients. We show that charge and thermoelectric transport in SWCNT networks can be modeled by the Boltzmann transport formalism, incorporating transport in heterogeneous media and fluctuation-induced tunneling. Considering the diameter-dependent one-dimensional density of states (DoS) of the SWCNTs composing the network, we can simulate the charge density and temperature-dependent Seebeck coefficients. Our simulations suggest that scattering in these networks cannot be described as simple one-dimensional acoustic and optical phonon scattering as for single SWCNTs. Instead the relaxation time is inversely proportional to energy (τ ∝ (E - EC)s, s = -1, EC being the energy of the first van Hove singularity), presumably pointing toward the more two-dimensional character of scattering events and the necessity to include scattering at the SWCNT junctions. Finally, our observation of higher power factors in trap-free, 1,2,4,5-tetrakis(tetramethylguanidino)benzene-treated (6,5) networks than in the RN networks emphasizes the importance of chirality selection to tune the width of the DoS. To benefit from both higher intrinsic mobilities and a large thermally accessible DoS, we propose trap-free, narrow DoS distribution, large-diameter SWCNT networks for both electronic and thermoelectric applications.

Влияние параметров сети на определение места повреждения при двухстороннем замере в фидере 35 кВ с трехобмоточным питающим трансформатором

Влияние параметров сети на определение места повреждения при двухстороннем замере в фидере 35 кВ с трехобмоточным питающим трансформатором

Different propagation speeds of recalled sequences in plastic spiking neural networks

Neural networks can generate spatiotemporal patterns of spike activity. Sequential activity learning and retrieval have been observed in many brain areas, and e.g. is crucial for coding of episodic memory in the hippocampus or generating temporal patterns during song production in birds. In a recent study, a sequential activity pattern was directly entrained onto the neural activity of the primary visual cortex (V1) of rats and subsequently successfully recalled by a local and transient trigger. It was observed that the speed of activity propagation in coordinates of the retinotopically organized neural tissue was constant during retrieval regardless how the speed of light stimulation sweeping across the visual field during training was varied. It is well known that spike-timing dependent plasticity (STDP) is a potential mechanism for embedding temporal sequences into neural network activity. How training and retrieval speeds relate to each other and how network and learning parameters influence retrieval speeds, however, is not well described. We here theoretically analyze sequential activity learning and retrieval in a recurrent neural network with realistic synaptic short-term dynamics and STDP. Testing multiple STDP rules, we confirm that sequence learning can be achieved by STDP. However, we found that a multiplicative nearest-neighbor (NN) weight update rule generated weight distributions and recall activities that best matched the experiments in V1. Using network simulations and mean-field analysis, we further investigated the learning mechanisms and the influence of network parameters on recall speeds. Our analysis suggests that a multiplicative STDP rule with dominant NN spike interaction might be implemented in V1 since recall speed was almost constant in an NMDA-dominant regime. Interestingly, in an AMPA-dominant regime, neural circuits might exhibit recall speeds that instead follow the change in stimulus speeds. This prediction could be tested in experiments.

Effect of network variables on the artificial neural network models for yield strength of austenitic stainless steel

PurposeThe purpose of the paper is to model the relationship between the yield strength of austenitic stainless steel and its chemical composition through the employment of artificial neural network (ANN). The effect of annealing temperature is also taken into consideration.Design/methodology/approachThe influence of network parameters, total number of neurons, number of neurons in a hidden layer, number of hidden layers and the interlayer distribution of neurons with a constant total neuron number, on the achievable training error is studied. Different learning rules available in MATLAB are used to assess the learning efficiencies of various networks.FindingsIt is found that increasing neuron number leads to a lowering of achievable training error up to a certain value beyond which training error remains constant. Increasing number of layers at constant total number of neurons causes a rise in the achievable training error. It is noted that if there is a sudden restriction of data flow in an ANN architecture, the achievable training error becomes higher. Interlayer distribution of neurons in ANNs used with different algorithms is found to have significant influence on the predictive performance of the networks.Originality/valueFrom the study on metallurgical validation of the output of various ANNs, it appears that mere attainment of a lower training error is not sufficient to achieve better generalization. A convergent network topology is found to be better than a divergent one in respect of effectively describing the input‐output relationship in austenitic stainless steel.

Protein‐Based Hydrogels with Tunable Dynamic Responses

AbstractDevelopment of hydrogel materials that respond to specific stimuli has been of significant interest in the design of modern functional materials. A variety of previous studies have used the ligand‐binding capability of proteins to design hydrogels that change their crosslinking density in response to stimuli. However, these materials generally undergo relatively small dynamic response, with limited control over response characteristics. This manuscript describes an alternative approach that exploits the ability of proteins to undergo nanometer‐scale conformational changes in response to stimuli. We report a class of novel protein‐based hydrogel materials that undergo tunable, reversible dynamic responses with a wide dynamic range (volume decreases to ∼25–90% of initial volume). These materials also undergo tunable, reversible changes in optical transparency (optical transparency decreases to ∼35–100% of initial optical transparency), and this phenomenon is used as a mechanism for label‐free biosensing. The materials are generated by photo‐crosslinking of an engineered version of the protein calmodulin flanked on each end with poly(ethylene glycol)‐diacrylate (PEGDA) moieties. The mechanism for dynamic changes derives from calmodulin's well‐characterized “hinge motion”‐upon‐ligand binding. Variations in network parameters in these hydrogels, including the molecular weight between cross‐links and the cross‐linking density, result in systematic tuning of material responsiveness. The influence of network parameters on hydrogel dynamics reported here may serve as a guide for the design of other protein‐based, responsive materials assembled using similar principles, and these materials may be a useful platform for design of biosensors, actuators, and drug delivery systems.

A Computational Modelling Approach to the Design of Stable Ventilation Systems

The well-known glove box arrangement used in the nuclear industry maintains a pressure depression within a contaminated volume while allowing passage of a small purge flow to the filtration plant. Conventionally, if a glove tears for example, an electromechanical control system restores the depression by opening and regulating an extra connection to the filtration duct. However, faster response and lower maintenance overheads result from replacing the conventional system with the fluidic vortex amplifier (VA) but coupled ventilation system and VA dynamics can be unstable. Designers needed a tool to predict oscillatory performance and, ideally, to advise on the strengths of influence of network parameters on stability. A system modelling approach is presented in which pipework dynamics, including steady state friction, are presented by transmission line modelling (TLM). A semi-empirical feedback model of linearized VA dynamics has been developed and shown to be substantially correct. The model's restriction to small signals does not adversely affect investigation of perturbation growth or decay. Partial transmission line representation of the VA model with special measures for feedback allowed time-domain simulations of network perturbations. However, this approach is indirect and it was possible to convert the whole VA model to transmission lines and then reformulate the TLM scheme into a direct, quantitative stability analysis. Extension of the technique provided a means of assessing the sensitivity of the stability to simple network alterations. The method has been programmed on an 80386 PC and successfully used on a large ventilation system. Success in the present context of dynamic stability for fluidic nuclear ventilation networks should be repeatable in other areas since the approach is applicable to any TLM-amenable system, be it fluid, mechanical, electrical or mixed.