To address the shortcomings of existing conveyor belt deviation detection methods, such as the small detection range and slow detection speed, a method that utilizes machine vision technology to detect belt deviations in belt conveyors is proposed. A track-type inspection robot was used to collect real-time running images of a conveyor belt; Wiener filtering was carried out, grayscale processing was used, Canny edge detection, Hough transform and least squares fitting operations were used to extract the straight lines of the edges of the conveyor belt, and the offset from the straight lines of the conveyor belt under normal running conditions was calculated. To address the cumulative error problem caused by the long time and long distance in the process of locating the wheel odometer, a joint localization technology that used radio frequency identification (RFID) and a wheel odometer was proposed, and the error correction was carried out for a rotary encoder by using the RFID technology. The experimental results demonstrate that the detection method achieved a detection accuracy of over 92.1% with a detection speed of up to 31 FPS. This method enables the fast and precise identification of conveyor belt deviations, and it satisfies the requirements for real-time detection for belt conveyors. The location accuracy of the localization system that combined RFID and a wheel odometer was able to reach the centimeter level, and the system could accurately locate the positions of belt deviations. This is of vital significance for ensuring safety and efficiency in production by enterprises.

: Strong noise interference can lead to failure of bearing fault diagnosis techniques. This paper proposes a two-step fault diagnosis strategy to address the challenge of weak fault feature extraction in bearing fault diagnosis using acoustic or vibration data at varying speed. Firstly, the paper introduces a short-time symplectic modal decomposition (stSGMD) method that utilizes fractional Fourier transform (FrFT). This method involves signal processing with short-time windowing to extract fault-sensitive components. The window is then expanded to obtain the complete component through fractional Fourier domain (FrFD) narrow-band filtering based on energy concentration in the FrFD. A novel entropy index, named standard deviation discrete entropy (SDDE), is introduced to quantify the intensity of fault shocks in non-stationary signal and is used to select components in the stSGMD. Subsequently, a fault feature extraction framework called global objective deconvolution (GOD) is presented for extracting instantaneous fault features at varying speed. This method establishes a global objective matrix for the extraction process. The GOD is utilized to deconvolute the complete fault-sensitive component, followed by envelope order analysis for demodulating the fault feature order. Numerical simulations and experimental studies on acoustics and vibration are performed. The results demonstrate that stSGMD improves the demodulation capability of SGMD, while GOD effectively extracts fault features. It is expected that the presented method will be effectively utilized for fault feature extractions in bearings operating under linear variable speed conditions.

Because it is difficult to extract multiple fault features from mechanical equipment under the interference of background noise and the parameters used in variational mode decomposition (VMD) must be determined in advance, a multiple fault separation method based on adaptive variational mode decomposition (AVMD) is proposed in this research to address these issues. Firstly, a novel index, known as the comprehensive impact coefficient (CIC), is established to properly identify the signal’s fault features. Thereafter, the fitness function of the sparrow search algorithm is developed based on the CIC, and the VMD parameters selection problem is solved. Finally, the decomposed modal components are subjected to envelop demodulation analysis, and the failure type of the bearing is assessed through the envelope spectrum. The simulation and experimental results reveal that the AVMD method can effectively separate all single faults from multiple faults, thus accurately diagnosing bearing faults.

Due to the slow detection speed, low accuracy, and small detection range of existing methods for detecting belt deviation in belt conveyors, this paper introduces an enhanced ultra-fast lane detection (UFLD) algorithm that leverages deep learning for the detection of belt deviation. Based on the UFLD algorithm, a variable step-size row anchor division method is proposed, and the simple parameter-free attention module is added to the network to enhance the network model’s focus on edge information of conveyor belts. Furthermore, improvements are made to the convolution operations in the ResNet-18 Stem and the downsampling operations in the residual modules, thereby enhancing the network’s ability to recognize the edges of conveyor belts. Based on the established experimental platform, a high-definition camera equipped with a track-type inspection robot was used to inspect the entire belt conveyor, covering the whole of the transmission line. The conveyor belt operation datasets collected under various working conditions were used to train and comparatively study the Hough Transform, DHT, YOLOv5, YOLOv8, LaneNet, SAD, and UFLD algorithms. The experimental outcomes demonstrate that the algorithm introduced in this article outperforms the other algorithms, achieving an F1-measure of 90.31%, an accuracy rate of 94.19 %, and a detection speed of 71 frames per second, meeting the real-time diagnostic needs for belt misalignment in the coal mining industry.

Broken steel wires in the core of the conveyor belt significantly reduce its carrying capacity and overall strength, potentially leading to safety hazards. The eddy current testing (ECT) has significant advantages in terms of cost and portability over other non-destructive testing technologies such as ultrasonic testing (UT) and X-Ray. ECT detects changes in the magnetic field, enabling the detection of the broken steel wires. However, the detection of broken wire ropes in conveyor belt has difficulty in obtaining data under actual working conditions, insufficient data diversity, low recognition accuracy, and poor model generalization ability. This paper designs a differential probe and introduces a novel approach, the CutMix-ReliefF-RF method, which addresses these challenges based on the differential signal. By applying CutMix, we enhance the diversity of small sample data and construct varied datasets representing different numbers of broken wires, which significantly improves the model’s recognition ability and generalization performance. The fault features of the broken steel wires in the core of the conveyor belt are extracted, then the ReliefF method is used for dimensionality reduction to obtain an optimal feature subset. Random forest (RF) algorithm is adopted to identify the broken core characteristics from the eddy current detection signal. The accuracy of the finite element model is validated through experimental and simulation signals used to generate training samples. The proposed method can accurately identify broken steel wires in the core of the conveyor belt, achieving a fault diagnosis accuracy of 98.23% under simulation signals and 96.67% under experimental signals, respectively. This provides strong support for the health monitoring of broken steel wires in the core of the conveyor belt.