Research on Feature Extraction and Intelligent Control of Molten Pool Based on Vision
|School||Nanjing University of Technology and Engineering|
|Course||Control Science and Engineering|
|Keywords||Arc Welding Robot Visual Sensing Molten Pool Image Near infrared Narrowband Filtering Visual Quality Characteristics Quality Intelligent Control Double Sight Vision Welding Procedure Intelligent Design|
Taking gas shielded arc welding robot process as research object,a research of molten pool(seam)vision near infrared auto-sensing technology, computer image processing,visual image of defects modeling and quality intelligent control is made in this paper.Based upon the process experiments on arc welding robot,the welding procedure auto design and on-line parameter sensing technologY and transmission system are researched.1.Welding Pool Visual Image Sensing with Near Infrared TechnologyFor spray,short circuit transfer and plasma heated molten pool,a method of using near infrared sensing to detect pool image on-line is presented.A breakthrough at the aspects of avoiding ghost image,near infrared narrowband and cut-off compoundfiltering,molten pool,seam and weld appearance visual integrative sensing is made.Someinnovated experiments on molten pool visual sensing Of plasma heating and deposited,aluminum alloy welding with twin-wire consumable electrode and plasma welding is made.Some clear images of molten pool of ptasma heating zone,TIG and deposited welding for copper,alloy welding with twin-wire consumable electrode are acquired.The key difficulty in visual sensing of quality control technology based on vision is resolved.2.Welding Pool Image Characteristics ExtractionSome image processing methods such as smoothing,Sobel sharpening,Cannny level operator,double floatation threshold auto-division,morphologic filtering,refinement,breakpoint connecting and image contour extraction is applied in the extraction of molten pool(seam)visual quality information. Aim at the problem of strong arc light disturbing the molten pool image characteristics,a speciai computer processing technology and software system of the molten pool(seam)visual image is developed and a clear molten pool (seam)geometry boundary,scum forming and quality characteristics is acquired.Combining the welding process parameters sensing and collection, an image processing method of disparting region quality characteristics is brought forward and the difficulty of molten pool contour and characteristics extraction is resolved.Based on above research,the describing method of molten pool(seam)visual image shape and characteristics is developed.And a new systemic parameter index for describing the molten pool(seam)visual image geometry shape is brought forward.The detailed indexes include molten pool area,drag angle,tail area,smooth grade of tail boundary,the relative symmetry degree of the molten pool vision image.3.Relation between Defects and Molten Pool Visual ImageBased on the molten pool image shape,boundary and scum characteristics, the molten pool image characteristics before occurring such defects as porosity,slag inclusion is researched for the first time.The main characteristics are as follows:the molten pool boundary showing as serration, some black flecks appearing at the molten pool tail,more scum,grey average value fluctuating and grey standard deviation becoming large.Those above characteristics may appear solely or several ones appear synchronously. Besides if those image characteristics appear,although it is not sure to bring porosity or slag inclusion,the probability of it will increase in great extent.4.Quality Intelligent Control Based on Visual SensingBased on fuzzy neuron net intelligent control technology,a fuzzy neuron net control model of weld width,penetration with spray,short circuit process based on visual sensing is also established.A further study on the experiments including disturbing factors is made and the corresponding model is corrected with experimental results.Results show that theestablished system has good effect on controlling the weld width.For the process which copper dePosited welding in TIG(plasma)arc,the quality auto-control method based on visual sensing is obtained;Some experiments including microstructure,electro microscope energy spectrum, high resolution transmission,mechanical performance testing are made.Results show that the eopper-steelboundary welding without penetration is obtained,the combination achieves atom grade and innovated results are obtained.5.Seam Molten Pool Double-Sight Visual Sensing TechnologyBased on the research of laser seam tracking,the laser vision seam geometry parameters sensing and tracking technology is developed.A structured laser double-sight visual sensing system is designed and studied.It.realizes the matching of Corresponding point in stereo-vision. The image Coordinate of seam characteristic point on two-dimension and geometry characteristics information such as groove type,angel,clearance are acquired.Using three-dimensional rebuilding arithmetic the three-dimensional coordinate can be calculated and the accurate three-dimensional geometry parameter of seams such as V、Half-V、butt-I、U and lap joint are acquired.6.Welding Procedure Auto-design System Software of Heavy Vehicle Gas Shielded Arc Welding RobotFor vehicle welding procedure intelligent design,a welding procedure auto-design system software of gas shielded arc welding(PMIG、MIG、MAG and SMAW)is presented and developed.The difficulty of double deducing based on the rules and experimental report welding procedure card,the Structured and semi-structured welding products information expressing and managing, information exchanging of welding CAD and proceduresystem figure is resolved. It also realizes the auto-resolving of the procedure with multi parameters and multi rules restriction in deducing process and batch procession output of products’ welding procedure.The software and hardware of real time sensing,issuing and remote transmitting of welding parameter for arc welding robot is designed and developed at thesame time.It realizes thesensing and cybernation transmission of welding current,arc voltage,welding speed and the moltenpool visual image.