Yiming Huang

Autonomous Detection of Weld Seam Profiles via a Model of Saliency-Based Visual Attention for Robotic Arc Welding

This paper presents a method of autonomously detecting weld seam profiles from molten pool background in metal active gas (MAG) arc welding using a novel model of saliency-based visual attention. First, a vision sensor based on structured light is employed to capture laser stripes and molten pools simultaneously in the same frame. Second, to effectively detect the weld seam profile from molten pool background for next autonomous guidance of initial welding positions and seam tracking, a model of visual attention based on saliency is proposed. With respect to the enhanced effect of saliency, the proposed model is much better than the classic models in the field. According to the comprehensive saliency map created by the proposed model, the weld seam profile can be extracted after threshold segmentation and clustering are applied to it in turn. Third, different weld seam images are used to demonstrate the robustness of the proposed methodology and last, to evaluate the performance of the proposed method, a measure called profile extraction rate (PER) is computed, which shows that the extracted weld seam profile can basically meet the requirements of seam tracking and the guidance of welding torches.

Parameter Self-Optimizing Clustering for Autonomous Extraction of the Weld Seam Based on Orientation Saliency in Robotic MAG Welding

This paper presents an effective method which needs free parameters as little as possible to autonomously extract the weld seam profile and edges from the molten background in two kinds of weld images within robotic MAG welding. First, orientation saliency detection produced by Gabor filtering nicely highlights the weld seam profile and edges from the molten background. Then, an unsupervised clustering algorithm combing a cluster validity index via an optimization rule, referred to as parameter self-optimizing clustering, is applied to discern the weld seam profile and edges from interference data after the orientation saliency detection result is given threshold segmentation. The validity index is better than the classical ones in two kinds of data sets through considerable tests. Last, two common applications of weld seam identification demonstrate the effectiveness of the proposed method.

The selection of arc spectral line of interest based on improved K-medoids algorithm

In order to eliminate the effect of wavelength error value and spectral line broadening on the definition of arc plasma spectrum, K-medoids algorithm is used to cluster different kinds of spectral lines and determine the spectral line of interest(SLOI). An improved K-medoids algorithm based on minimum spanning tree is proposed to solve the problem that K-medoids algorithm can not ascertain the number of classification. Moreover, spectral distance(SD) is proposed as the criterion to cluster in terms of the characteristic of spectral data. By marking the known spectral lines, cluster testing is made to validate the validity of the algorithm. The experiment results show that improved K-medoids algorithm can cluster effectively and determine the SLOI.

On the Mechanism and Detection of Porosity During Pulsed TIG Welding of Aluminum Alloys

Porosity is the remarkable barrier to realize the high efficient automatic welding of aluminum alloy. The paper provides new insights into the nucleation mechanism of porosity formation during pulsed TIG welding of aluminum alloys. Firstly, the model of bubble’s nucleation based on inclusions is proposed. The calculated pore minimum radius shows large sensitivity to pulse duty cycle as well as welding current. Then a novel method based on spectral analysis to detect porosity is developed. The relationships among the extracted signals and porosity defection are discussed, and the results show that Ar and HI spectral lines can be used as an aid to determine the most likely position of the porosity.

Weld penetration identification for VPPAW based on keyhole features and extreme learning machine

Variable polarity plasma arc welding, as an advanced manufacturing technology, has been successfully used in industrial production due to high energy density. The need for the control of the weld penetration remains of a long term interest in VPPAW process. In this study, a simple-flexible vision system was established to acquire a series of keyhole images, and the geometrical appearance of keyhole including the keyhole width and area are extracted based on part-based tree model. Then the acquired keyhole features are used to predict the weld penetration by using a novel extreme learning machine model. The research shows that ELM model can predict the penetration state of variable polarity plasma arc welding credibly and achieve real time monitoring for welding quality.

A Detection Framework for Weld Seam Profiles Based on Visual Saliency

To guide the accurate shift of welding robots in welding processes in real time, a newly designed visual sensor based on structure light is positioned on the torch in the robotic welding system. Through the sensor, one weld seam image includes the information of weld pools, laser stripe as well as the strong arc glare simultaneously. In order to accomplish the guiding task, a novel framework based on visual saliency is presented to detect the weld seam profile. Considerable image-processing experiments demonstrate the proposed framework is effective even in the background of lots of disturbances of fume, spatter and arc glare.