Laser Sensor Assisted Industrial Arc Welding Robot System for Producing Quality Welded Joints

Abstract

The task of automating the welding operation requires motion control, sensor integration and coordination with the welding power source. It is essential to set the robotic welding parameter specifications before starting of any welding cycle as these parameters affect the weld quality differently. The weld quality parameters are highly affected by the GMAW process parameters such as welding current, welding voltage, gas flow rate, robot travel speed and robot orientation etc. In the present research work, the effect of these process parameters on the weld quality has been studied combinely. The weld quality has been measured in a combined form of welding joint properties like bead penetration, width, height, ultimate strength, yield strength, micro-hardness, width of heat affected zone and surface defects.The prediction and optimization of this type of problem can’t be achieved by a single objective optimization technique as the input welding parameters affect the output parameters differently. Therefore, it is essential to introduce artificial intelligence technique in to the robotic welding process so that the best weld quality can be predicted from the manipulation of weld parameter setting without performing a number of experiments. A new approach based on combined Fuzzy and ANN (Artificial Neural Network) logic is presented for predicting best weld quality measures with respect to variations in weld process parameters. First, Fuzzy logic has been used to solve the complex interrelationships among the multiple weld quality measures like weld bead features, mechanical and microstructural properties by converting them in to a single Weld Quality Characteristic Index (WQCI). FIS can effectively acknowledge these conditions into internal hierarchy of it, by which it can overcome different drawbacks of other optimization techniques. Then ANN has been utilized to predict the WQCI from input of different settings robotic welding process parameters. From the results and analysis, it has been depicted that the developed neural network structure with four hidden layers can predict multiple weld quality measures from various welding process variables efficiently with a very strong correlation between the experimental and predicted outcomes by which cost involving for quality inspection in experimental trials can be removed. Afterthat, multi objective hybrid approaches like Fuzzy with meta-heuristic algorithms have been applied to achieve the optimal robotic welding conditions. Fuzzy-regression based approach has been utilized to establish mathematical model for the combined weld quality measures. Finally, an Enhanced TLBO algorithm has been applied to get optimal setting of the all the significant welding parameters. The proposed methodology has been found to be more efficient in achieving best weld quality in terms of maximized yield strength, ultimate strength, micro-hardness of welded joint, weld bead fautures with minimized heat affected zones, welding defects in comparison to other existing optimization techniques In this research work, robot assisted gas metal arc welding (GMAW) is introduced by controlling two subsystems; welding equipment and the laser sensor integrated industrial welding robot. An intelligent control strategy is adopted here to identify the feasible process related parameters and these are controlled in a way to obtain the desired weld quality. At first vision sensor has been used to obtain the seam positions, so that there is no need to teach the robot manually. For repeatable environment or for similar designed weld path, the weld seam can then be found by laser sensor and if there is any error any positioning, it is then notified to robot controller to move the robot end effector to the actual position. By this, if there is any change in weld gap in the weld seam, the robot tool centre point can be positioned accordingly. The output data of the sensor in terms of variation of weld gap has been used for control of welding parameters with Fuzzy controller to achieve maximum weld quality. This algorithm performs the seam finding in terms of weld geometry using laser sensing technology and alters the welding process parameters according to the geometric variations. Moreover, this paradigm maintains the torch orientation which is equipped to the end effector mounting plate in order to produce smoother and quality weld joints during the real time arc welding applications. The proposed Fuzzy controller based approach with laser and vision sensor intregration has been fond to be a cost effective, flexible method adding more value to the process of automatic robotic welding welding process. Use of robots in arc welding operations needs planning of trajectory path positions along which the robot end effector will be moved. For this, an optimal minimum time-jerk-acceleration-torque rate weld seam trajectory planning for MOTOMAN MA-1440 welding robot has been developed. This trajectory problem has been solved by newly developed hybrid Evolutionary Algorithm like NSGA-II with Achievement Scalaraizing Function local search algorithm to obtain a Pareto front which can give trade-offs between the objectives. From the simulation and experimental results, it can be concluded that by the implementation of the proposed optimal trajectory planning a smooth robot joint trajectory with reduced vibration of robot end effector travel, velocity stability and uniform weld bead along the weld seam can be achived. The influence of robot parameters like robot orientation, robot travel speed, focal length of sensor on the positional error, associated joint jerk-torque rate and the weld quality has also been investigated. NSGA-II with Nelder mead local search hybrid multi-objective algorithm has been proposed to obtain optimal values of robot and sensor parameters for getting minimum positional error with minimum jerk-torque rate and maximum weld quality. From the experimental investigation of the proposed approach, it has been observed that the measured weld quality features, positional errors of tool centre point along weld seam and joint kinematic parameter and dynmic parameter values have been optimized with the output setting of evolutionary algorithm over repeated cycle.