Research On Key Technologies Of Welding Staring Point Positioning And Obstacles Recognition Of Automatic Grinder Welding

With the deepening of China's intelligent manufacturing innovation strategy,the transformation and upgrade of traditional manufacturing industry are faced with enormous opportunities and challenges.As an indispensable material processing method in the traditional manufacturing industry,it has become an inevitable trend for welding to realize the intelligent upgrade and full automation.One of the most common welding structural parts in the heavy equipment industry,grinder is widely used in the fields of bridge,railway transportation,transportation container,lifting machinery,building and so on.Nowadays,due to the fact that random errors on vertical height and deflection angle caused by the manual operation exist in the production line when fixing and clamping displacement machine or in the central tilting process,it is difficult to realize self-positioning of welding starting point.At the same time,the random distribution of launders,reinforcing plates,upper auxiliary fenders and other obstacles that characterized with variable sizes and multiple structures on grinder workpiece makes it tough to achieve obstacles recognition and intelligent avoidance by traditional vision recognition algorithm.Aiming at several technical difficulties in full automation of grinder welding,this paper mainly researches on the positioning of welding staring point and obstacles recognition and clustering of automatic grinder welding by intelligent transformation,innovation and upgrade of the traditional semi automatic grinder production line.This paper presents a self-positioning method of welding starting point by composite-detection-typed dual eddy current positioning sensor.According to the influence factors and laws of eddy current sensor,the mechanism of how eddy current detects height and area is clarified.By setting up a double eddy current positioning system model,the variation law of complex impedance is obtained based on the calculation of inductance as well as equivalent analysis.A simple data acquisition test platform is built and the sampling strategy is optimized to get the output feature values by dual probe.The weighted least square method is adopted fitting the function of output feature values of probe 1 and height.Then variables can be separated based on the response surface methodology.Plugging the height values of probe 1,the fitting function of output feature values of probe 2 and the deflection angle is obtained.According to the scheme that positioning welding starting point by composite-detection-typed dual eddy current positioning sensor and the three main positioning feature parameters,the distancebetween the longitudinal and transverse directions of the welding torch is calculated.This paper presents a real-time cluster recognition of obstacles by broken line linear array laser sensor.The light path layout and parameters of broken line laser sensor can be designed referring to the principle of triangulation.The hardware circuit of laser sensor can be designed based on dsPIC30f4012 microcomputer.By analyzing the signal acquisition and imaging mechanism of laser vision sensors briefly,noise interference can be effectively suppressed using high-power lasers and specific attenuation slices and the effective length of line that adaptive to unified optimization scheme can be designed to ensure stable and reliable obstacle recognition.Specific to the defects of the traditional FCM applied to the real-time cluster recognition of obstacles on grinders,in this paper,it introduces real-time clustering strategy,replaces distance function and realizes global fast optimization to optimize fuzzy C-means real-time clustering algorithm.Through the comparison and analysis of the MATLAB platform,it is proved that the algorithm can acquire the number of clusters and the type attribute of obstacles in real time.According to results of positioning experiment of welding starting point on the automatic grinder welding simulated platform,it is quick in dynamic response,wide in application scope and precise in transverse and vertical position.Also,a real-time obstacles recognition experiment is carried out in the automatic grinder welding production line of a certain company and it proves that it is good in real time,accurate in clusters calculation and precise in obstacles avoidance.