As a result of long‐term pressure from train operations and direct exposure to the natural environment, rails, fasteners, other components railway track lines inevitably produce defects, which have impact on safety operations. In this study, multiobject detection method based deep convolutional neural network that can achieve nondestructive rail surface fastener defects is proposed. First, rails fasteners image are localized by improved YOLOv5 framework. Then, defect model Mask R‐CNN...

In complex indoor environments, traditional localization methods often suffer from non-line-of-sight (NLOS) and multipath problems, which lead to unsolvable or incorrectly solved mathematical models, thereby limiting accuracy. A method based on swarm intelligence optimization has been proposed address this issue. The algorithm does not require solving matrix inversions transforms the problem into a function problem, can obtain approximate optimal solutions. Nevertheless, algorithms are beset...

The total focusing method (TFM) is an ultrasonic phased array imaging algorithm used in nondestructive testing (NDT) that processes large amounts of data from full matrix capture (FMC). This limits its application some industrial fields with real-time requirements. To solve this problem, a sparse optimization applied to FMC-TFM can reduce time consumption and improve efficiency. However, conventional intelligent methods, such as genetic (GA), use binary encoding, which require intensive...

Railway track gauge irregularity severely reduces the service life of rail and vehicle, even result in vehicle falling off or wheel trapping, which causes driving accidents. In this paper, a dynamic inspection method based on computer vision is presented. According to method, system could be constructed by using four CCD (Charge-coupled Device) cameras two red laser sector lights. The principle corresponding calibration are analyzed. order get points, several image processing technologies,...

Using sparse matrix for total focusing method (TFM) imaging can improve computational efficiency in non-destructive testing (NDT). In design, binary particle swarm optimisation (BPSO) and genetic algorithm (GA) often fall into local optimal solution, the decreases significantly with expansion of array scale. this paper, a discrete war strategy (DWSO) is proposed to realise ultrasonic imaging. This uses equidistant scatter mapping real-number encode position limit search range. Then, fitness...

The operational reliability of rail vehicle pantograph systems is evaluated by transforming T-S multistate fault trees into dynamic Bayesian networks (DBNs), which take account system multistability, long-lasting operation, failure, and maintenance recovery. tree structure constructed the content validity ratio index; gate rule expressing causal uncertainty using fuzzy theory dependent uncertain ordered weighted averaging expert scoring, finally, transformed a DBN model characterizing...

Future developments in image guided adaptive radiotherapy (IGART) for bladder cancer require accurate deformable registration techniques the precise assessment of tumor and motion deformation that occur as a result large volume changes during course treatment. The aim was to employ an extended version point-based algorithm allows control over tissue-specific flexibility combination with authors' unique patient dataset, order overcome two major challenges registration, i.e., difficulty...

Electricity load forecasting is becoming one of the key issues to solve energy crisis problem, and time-series Bayesian Neural Network popular method used in forecast models. However, it has long running time relatively strong dependence on weather factors at a residential level. To these problems, this article presents an improved Networks (IBNN) model by augmenting historical data as inputs based simple feedforward structure. From delays correlations impact analysis, containing different...

Purpose: The aim of this study is to develop and validate a generic method for automatic bladder segmentation on cone beam computed tomography (CBCT), independent gender treatment position (prone or supine), using only pretreatment imaging data. Methods: Data 20 patients, treated tumors in the pelvic region with entire visible CT CBCT, were divided into four equally sized groups based position. full empty contour, that can be acquired imaging, used generate patient‐specific shape model. This...

This paper aims to use the Lamb wave local wavenumber approach characterize flat bottom defects (including circular holes and a rectangular groove) in an isotropic thin plate. An air-coupled transducer (ACT) with special incidence angle is used actuate fundamental anti-symmetric mode (A0). A laser Doppler vibrometer (LDV) employed measure out-of-plane velocity over target area. These signals are processed by domain filtering technique order remove any modes other than A0 mode. The filtered...

This paper aims to create a prediction model for car body vibration acceleration that is reliable, effective, and close real-world conditions. Therefore, huge amount of data on railway parameters were collected by multiple sensors, different correlation coefficients selected screen out the closely correlated acceleration. Taking previous as inputs, was established based several training algorithms neural network structures. Then, successfully applied predict test datasets segments same...

The Pandrol track fastener image is composed of two parts: clip sub-graph and bolt sub-graph. However, the detection defect can be realized by cannot effectively detect whether loose. When convolutional neural network used to extract whole picture features detect, many unrelated clips will obtained, thereby resulting in a high false alarm rate. To solve these problems, method based on local defects proposed. First, algorithm for automatic segmentation pictures was divide into sub-pictures,...

The current Lamb wave topology imaging methods are limited to the ideal model of through-hole defects, ignoring influence mode coupling. However, blind hole such as structural plate corrosion and cracks, most likely defects in practical applications, waves bound appear mode-coupled at these defects. use full matrix data calculations direct concomitant sound fields is time-consuming limits industrial application ultrasonic time-domain topological energy (TDTE) method. A TDTE method a sparse...

A method of combining Green’s function retrieval theory and ultrasonic array imaging using Lamb waves is presented to solve near filed defects in thin aluminum plates. The are close the phased satisfy field calculation formula. Near acoustic information obscured by nonlinear effects initial wave signal a directly acquired response full matrix capture mode. reconstructed inter-element responses produced from cross-correlation received signals between sensor pairs. This new eliminates...

Commonly used fastener positioning methods include pixel statistics (PS) method and template matching (TM) method. For the PS method, it is difficult to judge image segmentation threshold due complex background of track. TM search in both directions global easily affected by background, as a result, locating accuracy fasteners low. To solve above problems, this paper combines with proposes new called local unidirectional (LUTM). First, rail achieved based on gray-scale vertical projection....

This work presents a new method for sleeper crack identification based on cascade convolutional neural network (CNN) to address the problem of low efficiency and poor accuracy in traditional detection identification. The proposed algorithm mainly includes improved You Only Look Once version 3 (YOLOv3) recognition network, where two modules, encoder-decoder (CEDNet) residual refinement (CRRNet). YOLOv3 is used identify locate cracks sleepers segment them after ballast bed extracted by using...

In this paper, the vehicle faults are estimated and diagnosed by introducing wavelet transform, based on state-space method.It monitors state of railway suspension system, establishes a vertical dynamic model railway-vehiclesystem to identify parameters system. The simulation results show that, in circumstances thechange which is resulted from gradual fault or composite simulationresults can effectively basic characteristics its parameters, realize diagnosis, so as achieve thepurpose...

Loosening detection; cascade deep convolutional neural network; object localization; saliency detection problem of bolts on axlebox covers. Firstly, an SSD network based ResNet50 and CBAM module by improving bolt image features is proposed for locating And then, the A2-PFN according to slender marker lines extracting more accurate regions bolts. Finally, a rectangular approximation method regularize line as way calculate angle plot all values into table, which criteria table can determine...

In the track irregularity detection, acceleration signals of inertial measurement unit (IMU) output which with low frequency components and noise, this paper studied a de-noising algorithm. Based on criterion consecutive mean square error, method for IMU based empirical mode decomposition (EMD) was proposed. This can divide intrinsic functions (IMFs) derived from EMD into signal dominant modes noise modes, then reflecting important structures were combined together to form partially...

In this paper, an EMD de-noising algorithm is proposed based on the statistical feature of random noise, which can eliminate noise impaction digital integrator generated by collected railway line state detection signals using strap-down inertial technology. Firstly, first IMF component noise-dominant modes treated process “random sort-sum-average-reconstruc-tion”, signal-to-noise ratio improved while power weakened in process. Then cut-off be determined according to characters...