Forming Characteristics In Multi-layer Single-bead GMA Additive Manufacturing And Control For Deposition Dimension

Gas metal arc (GMA) additive manufacturing (AM), using welding arc as theheat source and wire as the deposition material, fabricates metal parts by layersbased on3D models. At present, most of the works for GMA-AM are limited to thefeasibility of the technology, microstructure, and mechanical properties of parts,little has been done for process stability, surface quality, and dimension accuracy.This paper focuses on the process characterisitcs, surface quality, online detectionand closed loop control of multi-layer single bead GMA-AM, with the aim ofenhancing the process stability and formed dimension accuracy.The effects of process parameters on forming appearance of the single-layersingle-bead were studied, and the process specification interval for excellentappearance was obtained. The influences of arc current, deposit velocity, and heatinput on forming appearance of multi-layer single-bead were investigated. Theresults showed that the current is the major influence factor of forming appearance.As the current was larger than200A, serious instability and overflow of pool wereproduced from second to fourth layers, and the forming appearance of the partcollapsed. The optimum current ranged between100A and180A. With the increasein current, the maximum heat input for good appearance decreased.A second-order regression analysis, using a central composite rotatable design,was used to establish the relationship between the process parameters and layergeometry. This model can predict the deposited layer width and height with differentprocess parameters, providing bases for slices of the3D model.According to the characteristics of high and wide arc striking points, and lowarc extinguishing points, corresponding control strategies were proposed. Forclosed-path parts, arc striking and extinguishing points in the same layer weresuperposed to make up for height difference. For open-path parts, one strategy wasto use the alternating deposition direction in adjacent layers. The other strategy wasto use the same deposition direction. Arc striking points increased deposit velocity,and decreased the ratio of wire feed speed to deposit velocity (RWFSDV). Arcextinguishing points decreased current, deposit velocity, and increased RWFSDV.Influence factors of surface quality for multi-layer single-bead GMA-AM werefurther studied. A laser vision sensor was utilized to detect the surface quality, andassessment system for surface quality was established. The effects of inter-layertemperature, constant RWFSDV, and variable RWFSDV on surface quality werestudied. The surface quality is related to effect of multi-factor. With the sameprocess parameters, the higher the inter-layer temperature, the worse the surface quality. With RWFSDV remaining constant, the current increases, the surfacequality decreases. Keeping the current and inter-layer temperature as constant, theincrease in deposit velocity will result in the increase of the surface qualit y. Keepingthe deposit velocity and inter-layer temperature as constant, the increase in currentwill result in the decrease of the surface quality.For real-time detection of deposition dimension geometry, a double passivevision sensor system was designed for dimension control. The system can obtainimages of layer width and the nozzle to top surface distance (NTSD). Imageprocessing algorithms, consisting of Gaussian filter, Sobel operator, and Houghtransformation, were applied to extract the characteristic points of the beadgeometry for real-time detection, which will lay a foundation for process modelingand controller design.The dynamic single-input single-output models between process parametersand deposition dimension were acquired by step responses, and the time-domainanalysis of the system was done. The forming process showed non-linearity andtime-delay. The simple model can not describe the dynamic relationship. Usingrandom experimental data, the dynamic relationships were identified by means of anon-linear Hammerstein model, which was used for controller simulation.A single neuron self-learning proportional summational differential (PSD)controller was designed for layer width control. The controlling variable was thedeposit velocity. Simulation results and disturbance experiments validated theperformance of the controller. The PSD controller can be effectively applied to theconstant and variable width control. In order to meet the stability of the depositionheight, the NTSD was also monitored, and an adaptive controller was designed forNTSD control.A simulated combustion chamber shell of airspace engine, a circular casing ofaeroengine, and a thrust chamber of rocket engine were fabricated by GMA-AMtechnology. The parts showed excellent surface quality. The vision sensor systemand PSD control results confirmed that the control accuracy of the layer width wasless than0.4mm.