This paper presents a convolutional neural network (CNN) based approach for fault diagnosis of rotating machinery. The proposed incorporates sensor fusion by taking advantage the CNN structure to achieve higher and more robust accuracy. Both temporal spatial information raw data from multiple sensors is considered during training process CNN. Representative features can be extracted automatically signals. It avoids manual feature extraction or selection, which relies heavily on prior...

The degradation of bearings plays a key role in the failures industrial machinery. Prognosis is critical adopting an optimal maintenance strategy to reduce overall cost and avoid unwanted downtime or even casualties by estimating remaining useful life (RUL) bearings. Traditional data-driven approaches RUL prediction rely heavily on manual feature extraction selection using human expertise. This paper presents innovative two-stage automated approach estimate deep neural networks (DNNs). A...

Predictions of renewable energy (RE) generation and electricity load are critical to smart grid operation. However, the prediction task remains challenging due intermittent chaotic character RE sources, diverse user behavior power consumers. This article presents a novel method for using improved stacked gated recurrent unit-recurrent neural network (GRU-RNN) both univariate multivariate scenarios. First, multiple sensitive monitoring parameters or historical consumption data selected...

Intelligent fault diagnosis techniques play an important role in improving the abilities of automated monitoring, inference, and decision making for repair maintenance machinery processes. In this article, a modified stacked autoencoder (MSAE) that uses adaptive Morlet wavelet is proposed to automatically diagnose various types severities rotating machinery. First, activation function utilized construct MSAE establish accurate nonlinear mapping between raw nonstationary vibration data...

This paper proposes a new method for stability analysis of bidirectional dual full-bridge dc-dc converter with triple phase-shift control under arbitrary parameter changes. The present makes the determination these power converters more systematic and precise than existing methods in this field, which are largely based on simulation. Nonlinear periodic operation is presented including circuit. Using working theory, separated into several stages. Equivalent circuits state equations developed...

Learning on point cloud is eagerly in demand because the a common type of geometric data and can aid robots to understand environments robustly. However, sparse, unstructured, unordered, which cannot be recognized accurately by traditional convolutional neural network (CNN) nor recurrent (RNN). Fortunately, graph (Graph CNN) process sparse unordered data. Hence, we propose linked dynamic CNN (LDGCNN) classify segment directly this paper. We remove transformation network, link hierarchical...

Robots are increasingly operating in indoor environments designed for and shared with people. However, robots working safely autonomously uneven unstructured still face great challenges. Many modern wheelchair accessibility mind. This presents an opportunity wheeled to navigate through sloped areas while avoiding staircases. In this paper, we present integrated software hardware system autonomous mobile robot navigation environments. modular reusable framework incorporates capabilities of...

Visual information is indispensable to human locomotion in complex environments. Although amputees can perceive the environmental by eyes, they cannot transmit neural signals prostheses directly. To augment human–prosthesis interaction, this article introduces a subvision system that environments actively, assist control powered prosthesis predictively, and accordingly reconstruct complete vision-locomotion loop for transfemoral amputees. By using deep learning, classify common static...

Camera relocalization plays a vital role in many robotics and computer vision tasks, such as global localization, recovery from tracking failure, loop closure detection. Recent random forests based methods directly predict 3D world locations for 2D image to guide the camera pose optimization. During training, each tree greedily splits samples minimize spatial variance. However, these greedy often produce uneven sub-trees training or incorrect 2D-3D correspondences testing. To address...

Environmental monitoring is a practical application where wireless sensor network (WSN) may be utilized effectively. However, the energy consumption issues have become major concern in using WSN, particularly remote locations without readily accessible electrical power supply. In general, activities of data transmission among nodes and gateway (GW) can significant fraction total within WSN. Hence, reducing number duration transmissions as much possible while maintaining high level accuracy...

This letter presents a novel and integrated framework for Next-Best-View (NBV) selection toward autonomous robotic exploration in indoor environments. A topological map, named semantic road map (SRM), is proposed to represent the explored environment during exploration. The basic concept of SRM construct graph with nodes containing states edges satisfying collision-free constraints. Especially, integrates both structure information environment, which possesses beneficial properties using It...

Current fault diagnosis methods for rotor-bearing system are mostly based on analyzing the vibration signals collected at steady rotating speeds. In those methods, data under one operating condition cannot be accurately used a different condition. Moreover, in monitoring, installing necessary sensors will affect equipment structure and hence response itself. The present paper proposes new method two-stage parameter transfer infrared thermal images of variable enables use (or parameters)...

This paper concerns energy harvesting from vehicle suspension systems. The generated power associated with bounce, pitch, and roll modes of dynamics is determined through analysis. potential values generation these three are calculated. Next, experiments carried out using a four jack shaker rig to validate the analytical harvest. For considered vehicle, maximum theoretical 1.1, 0.88, 0.97 kW modes, respectively, at 20 Hz excitation frequency peak-to-peak displacement amplitude 5 mm each...

In this article, the problem of cross-domain fault diagnosis rotating machinery is considered. a practical setting approach, operating platform machine may have different setup and conditions compared to experimental that used collect training data. This can lead significant data variations, specifically domain shifts. Conventional data-driven approaches are known adapt poorly these shifts, resulting in drop accuracy when pretrained model applied actual situation. an unsupervised adaptation...

Learning through a point cloud is attractive be-cause contains geometric data and can help robots understand environments in robust manner. However, sparse, unstructured unordered, its accurate recognition by traditional convolutional neural network (CNN) or recurrent (RNN) difficult. Hence, the present paper proposes linked dynamic graph CNN (LDGCNN) to directly classify segment cloud. The work removes transformation network, links hierarchical features from graphs, freezes feature...

Real-time motor fault diagnosis can detect faults on time and prompt the repair or replacement of faulty motors which minimizes potential losses caused by faults. Deep learning (DL) methods have been intensively applied in diagnosis. Most DL algorithms need to be trained with sufficient computation resources cloud local servers. However, uploading raw data downloading command instructions edge will cause inevitable delays security concerns. This paper develops a algorithm based efficient...