Human Welder's Response-based Interactive Behavior Analysis Of Welder With Weld Pool And Control Of Weld Pool Morphology

In a complex and varying in-situ welding condition,automatic welding robot should be equipped with two high abilities like a skilled human welder,including the good ability of weld seam tracking and path planning and the effective control ability of weld pool dynamics for replacing the partial or more human welder's works.Currently,the reasonable weld seam tracking issue can be primely realized using advanced sensors and intelligent path planning algorithms.However,the weld metal fulid flow can not be well controlled by the robot since the controled objective of weld pool is fuzzy and divergent.Hence,how to stabilize and control the weld metal fluid flow becomes a very important factor to develop the next intelligent welding technology and improve the adaptability of the welding robot ultilized in the complex welding conditions.To this end,this paper tried to learn and mimic the adjustment of the skilled human welder in the weld pool control process,and then understand and extract the control rule of the welder as well as model this interactive behavior of human welder with weld pool surface.To complete above goal,a new scheme that models the welder's reaction and controls of the weld pool morphology was proposed based on the weld pool surface which was regarded as the interactive interface between the human welder and weld pool dynamics.Numerical and experimental investigations verified the following conclusions drawing from this work.(1)To visualize the interactive process of human welder and weld pool and investigate the characterizations of human welder's response on the varying weld pool surface,a laser-vision based experimental system combined a welding torch orientatin measurement using an inertial measurement unit(IMU)was established,and the data pairs including the welding torch and weld pool characteristic parameters were sampled synchronously.A quaterion-based Unscented Kalman Filter(UKF)algorithm coupled with an auto-nulling algorithm was developed to reduce the zero drift error of the gyrocope and estimate the welding torch orientation,.Meanwhile,an improved three dimensional reconstruction algorithm was adopted to reconstructe the weld pool surface.To describe the weld metal fulid flow behavior,the curvature radius of reflected laser stripes was proposed and achieved using a special image processing algorithm and curvature computing eqations.Finally,the correlation between the welding troch attitude and weld pool surface behavior was comparatively analyzed,indicating that the human welder not only adjusts the thermal distribution of weld pool surface but also adjusts the forc balance acting on the weld pool.This interactive action completely represents the thermodynamics adustments and shows the welder's experience and operating skills.The weld pool surface characteristic parameters reflect the comprehensive estimation and adjustments of weld pool dynamics made by welder.(2)A couple model of the moving weld pool frame and the welding torch frame was established to realize them synchronization in space-time field.Moroever,the effect rule of the welding torch orientation adjusted by different level skilled welder on the weld pool status was experimentally studied,and the results show that the novice welder irregularly and randomly adjusts the welding torch and lacks the basic welding skills and the awareness interactived with the weld pool surface;the desired weld appearance cannot be achieved.The less skilled welder's adjustments represent partially regular and multiparameters control characteristics,and its variation ranges in 2 degree.But,he/she cannot keep this regular adjustment in the whole welding process and lacks the global observation and control ability.Skilled human welder can control the weld metal fulid flow status based on his/her obaservations and pass-trained experience,representing a sensitive awareness of indentifying the variation of weld pool status.(3)Based on the analysis of experiments,the adjustment of welding torch on weld pool dynamics was numerically studied.The results indicate that the welding torch changes the energy distribution on the weld pool surface and the force direction acting on the weld pool,which in turn changes the weld metal flow behavior and finally represents it on the varying weld pool morphology.The numerical results also elaborate that the essential of the human welder's responses/experience is to change the thermodynamics coupled ablalnce and maintain the optimum weld metal fluid flow status.The obtained dynamic responsive time of weld pool provides a theoretical reference data for designing control system and optimizing the control algorithm to achieve the desired weld pool.(4)When the physical process that human welder adjusts the weld pool dynamic behavior is understand and characterized,an adaptive kernel radial basis function neural networks(AK-RBFNN)model was identified using data-driven approach.A humanmachine cooperative virsualized welding platform was developed to intelligently controled the weld pool dynamics under changing the welding current and the welding speed conditions.Experimetnal results show that the proposed AK-RBFNN model can better infuse the corresponding factors of human welder's response and distill the local details of the human welder adjustment process,and adaptively control the weld pool topography,producing a sound weld bead and realizing the intelligent welding process control under certain conditions.This results lay a signifincat foundation for improving the adaptive capacity of the welding robot in the complex in-situ welding conditions.