Search

Deep learning models have been widely used in data-driven bridge structural damage diagnosis methods in recent years. However, these methods require training and test datasets to satisfy the same distribution, which is difficult to satisfy in practice. Domain adaptation transfer learning is an efficient method to solve this problem. Most of the current domain adaptation methods focus on close-set scenarios with the same classes in the source and target domains. However, in practical applications, new damage caused by long-term degradation often makes the target and source domains dissimilar in the class space. For such challenging open-set scenarios, existing domain adaptation methods will be powerless. To effectively solve the above problems, an adversarial auxiliary weighted subdomain adaptation algorithm is proposed for open-set scenarios. Adversarial learning is introduced to proposed an adversarial auxiliary weighting scheme to reflect the similarity of target samples with source classes. It effectively distinguishes unknown damage from known states. This paper further proposes a multi-channel multi-kernel weighted local maximum mean discrepancy metric (MCMK-WLMMD) to capture the fine-grained transferable information for conditional distribution alignment (sub-domain alignment). Extensive experiments on transfer tasks between three bridges verify the effectiveness of the algorithm in open-set scenarios. [ABSTRACT FROM AUTHOR]

In this article, a novel yet simple step climbing robot is proposed and is comprised of two front wheels, a rear-wheel and an actuator to vary the center distance between the front and rear wheels. When a robot climbs a stair, the huge variance in the inclination angle of the robot may result in its toppling. Hence, a mechanism is proposed to compensate for the change in inclination angle. An inertial measuring unit (IMU) is used to sense the inclination angle of the robot which is then fed to a microcontroller in order to actuate the connecting link, thereby reducing the variation of the inclination angle. During ascending simulations on dynamic model based on the Newton–Euler formulation, the required torque on rear wheel is reduced by 26.3% as compared to uncontrolled simulations. Moreover, the normal reaction on rear wheel during descending simulation has increased by 170.9% by controlling the inclination angle, which reduced the probability of toppling of the proposed robot. Multiple experiments on the prototype with a controlled condition show that the variation in inclination angle is reduced by 77.8% during ascending, whereas it is reduced by 92.8% during descending resulting in successful operation on the stairs as compared to uncontrolled cases. [ABSTRACT FROM AUTHOR]

Feature Pyramid Network (FPN) builds a high-level semantic feature pyramid and detects objects of different scales in corresponding pyramid levels. Usually, features within the same pyramid levels have the same weight for subsequent object detection, which ignores the feature requirements of different scale objects. As we know, for most detection networks, it is hard to detect small objects and occluded objects because there is little information to exploit. To solve the above problems, we propose an Enhanced Feature Pyramid Object Detection Network (EFPN), which innovatively constructs an enhanced feature extraction subnet and adaptive parallel detection subnet. Enhanced feature extraction subnet introduces Feature Weight Module (FWM) to enhance pyramid features by weighting the fusion feature map. Adaptive parallel detection subnet introduces Adaptive Context Expansion (ACE) and Parallel Detection Branch (PDB). ACE aims to generate the features of adaptively enlarged object context region and original region. PDB predicts classification and regression results separately with the two features. Experiments showed that EFPN outperforms FPN in detection accuracy on Pascal VOC and KITTI datasets. Furthermore, the performance of EFPN meets the real-time requirements of autonomous driving systems. [ABSTRACT FROM AUTHOR]

This manuscript presents a shape memory alloy (SMA) actuated anti-slip mechanism for the wheels of step climbing robots. The proposed mechanism comprises three kinematic chains considering the Lazy Tong and the bi-stable four-bar mechanism. Chain 1 of the mechanism is used to clamp on the edges of the stairs to avoid slipping. The second chain of the mechanism is used to switch the mechanism between two stable positions, i.e., open position and closed position, of chain 1. For activating the mechanism, the third chain is employed which is based on SMA wire. Furthermore, the mechanism is designed to achieve passive switching from the open position to the closed position. Equations are developed to determine the dimensions of various members. Using those parameters, a 3D model of the proposed mechanism is developed. Stress analysis is performed and the model is found to be safe under a load of 250 N with a factor of safety of 3.025. The mechanism is attached to either side of a wheel of the outer radius of 290 mm. To analyze the kinematics of the mechanism, a three-dimensional model in MSC Adams is developed and studied. The force required by SMA actuator is found to be less than 5 N. The proposed mechanism may be used for various unmanned robotic systems while mitigating step-like obstacles in the path. [ABSTRACT FROM AUTHOR]

For a system such as the concrete structure, flow can be the dynamic field to describe the motion, interactions, or both in dynamic or static (Eulerian description) states. Further, various kinds of flow propagate through it from the very start to the end of its lifecycle (Lagrangian description) accompanied by rains, winds, earthquakes, and so forth. Meanwhile, damage may occur inside the structure synchronously, developing from micro- to macro-scale damage, and eventually destroy the structure. This study was conducted to clarify the content of flow which has been implicitly used in the damage detection, and to propose a flow analysis framework based on the combination data space and the theory of dissipative structure theory specifically for nondestructive examination in structural damage detection, which can theoretically standardize the mechanism by which flow characteristics vary, the motion of the structure, or the swarm behavior of substructures in engineering. In this paper, a destructive experiment (static loading experiment) and a following nondestructive experiment (impact hammer experiment) were conducted. According to the experimental data analysis, the changing of flow characteristics shows high sensitivity and efficient precision to distinguish the damage exacerbations in a structure. According to different levels of interaction (intensity) with the structure, the information flow can be divided into two categories: Destructive flow and nondestructive flow. The method used in this research is named as a method of "flow analysis based on flow characteristics", i.e., "FC-based flow analysis". [ABSTRACT FROM AUTHOR]