Yanming Li

Robust Visual Monitoring of Machine Condition with Sparse Coding and Self-Organizing Map

A direct way to recognize the machine condition is to map the monitored data into a machine condition space. In this paper, via combining Sparse Coding and Self-Organizing Map, a new model (SCSOM) is proposed for robust visual monitoring of machine condition. Following the model, a Machine Condition Map (MCM) representing the machine condition space is formulated offline with the historical signals; then, during the online monitoring, the machine condition can be determined by mapping the monitoring signals onto the MCM. The application of the SC-SOM model for bearing condition monitoring verifies that the bearing condition can be correctly determined even with some disturbances. Furthermore, novel bearing conditions can also be detected with this model.

ARM based load and hook measuring and tracking for precision hoist of tower crane

It is important to measure the activities of the hook and the load for precision hoisting and safe operation of a tower crane. Visual monitoring and image recognition are the optimum methods for crane hook activity measuring and precision hoisting. Advanced RISC Machines (ARM) is a promising controller for field measuring and controlling of tower crane, but high real time performance and low computation requirements are required for measuring system implemented using ARM. An ARM-based load and hook measuring and tracking system for tower crane is designed. Selecting lifting rope as the target object, an high-performance Improved Progressive Probabilistic Hough Transform (IPPHT) algorithm is proposed for hook activity measuring and tracking. Compared to Progressive Probabilistic Hough Transform (PPHT) more accurate detection is obtained, and the same time the computation time can be reduced to 20%. It is tested the IPPHT is fitting for ARM based measuring system.

An evolving machinery fault diagnosis approach based on affinity propagation algorithm

An evolving approach combining unsupervised clustering and supervised classification for intelligent machinery fault diagnosis is proposed. As the key point of the approach, the unsupervised clustering module for detecting a novel fault is specified in this paper. Incorporating with prior information in the historical data, a constrained clustering method is developed based on the affinity propagation (AP) algorithm. The clustering method is validated through experimental case studies on the bearing fault diagnosis. The results show that the clustering method has the self-learning abilities to detect the novel faults and with the evolving abilities of the proposed approach, one can start with just the normal condition data and continue building the diagnosis scheme as the new fault events occur.

Rapid Developing the Simulation and Control Systems for a Multifunctional Autonomous Agricultural Robot with ROS

Building customized control system for specific robot is generally acknowledged as the fundamental section of developing auto robots. To simplify the programming process and increase the reuse of codes, this research develops a general method of developing customized robot simulation and control system software with robot operating system (ROS). First, a 3D visualization model is created in URDF (unified robot description format), and is viewed in Rviz to achieve motion planning with MoveIt! software package. Second, the machine vision provided by camera driver package in ROS enables the use of tools for image process, 3D point cloud analysis to reconstruct the environment to achieve accurate target location. Third, the communication protocols provided by ROS like serial, Modbus support the communication system development. To examine the method, we designed a tomato harvesting dual-arm robot, and conducted farming experiment with it. This work demonstrates the advantages of ROS when applied in robot control system development, and offers a plain method of building such system with ROS.

CMAC-based real-time calculation of the effective welding current during A.C. resistance spot welding

Real-time measurement of effective current is essential for fine control of welding process. However, conventional methods cannot simultaneously satisfy the accuracy and the real-time requirements. On the basis of the welding transformer circuit model, a new measuring method is proposed. This method measures the peak value of the welding current and then multiplies a coefficient K to obtain the root-mean-square value, which enables the processor to calculate the half-wave effective current at the moment of peak current. A Cerebellar Model Articulation Controller(CMAC) model is off-line trained and used online to compute K, which is a nonlinear function against the firing angle and the peak current angle. A DSP-based resistance spot welding monitoring system is developed to perform CMAC computation. Experimental results suggest that this measuring method is feasible.

Far-Field terrain perception using max-margin markov networks

Far-field terrain perception plays an important role in performing outdoor robot navigation, such as earlier recognition of obstacles, efficient path planning. Stereo vision is an effective tool to detect obstacles in the near-field, but it cannot provide reliable information in the far-field, which may lead to suboptimal trajectories. This can be settled through the use of machine learning to accomplish near-to-farlearning, in which near-field terrain appearance features and stereo readings are used to train models able to predict far-field terrain. In this paper, we propose a near-to-far learning method using Max-Margin Markov Networks (M3N) to enhance long-range terrain perception for autonomous mobile robots. The method not only includes appearance features as its prediction basis, but also uses spatial relationships between adjacent parts. The experiment results show that our method outperforms other existing approaches.