The Real-time Monitoring Research Of Molten Pool Shape During Laser Additive Manufacturing

Laser metal additive manufacturing is an innovative manufacturing technology derived from laser cladding technology and rapid prototyping technology.Additive manufacturing offers great flexibility for design and eases manufacture difficulties to produce geometrically complicated functional components.It fits the future scenario of low-volume customized production and high-added-value components repair and direct manufacture.Melt pool geometry,melt pool temperature,deposition height,and optical emission have been the parameters that are generally monitored during an additive manufacturing process.Melt pool geometry,as one of the most significant built attributes,reflects the process stability and the build quality,and has been widely used for closed loop control of an additive manufacturing process.And the observation of the melt pool is strongly disturbed by spatters,flares,non-fused or incandescent suspended powders,hot particles,oxidized surface layers,plasma and plume vapors,and melt pool surface disturbances.Melt pool geometry can be monitored by direct observation with proper filters,by observation of the reflected light with the illumination of a narrow band light source,or by analysis of a thermal image.Apart from narrow band filters to reduce the above mentioned noises and disturbances,many image processing algorithms,including fixed gray-level threshold method,f'uzzy entropy threshold method,Otsu method,and Canny method,are exploit to obtain the accurate melt pool geometry.All above melt pool image processing methods are based on gray-level images,and a fixed gray-level threshold method is mostly used.However,the image gradient values that correspond to significant edges are usually determined empirically.It also still has limitation to deal with randomly located blurred areas caused by unstable plasma plume as well as dark areas/speckle noise caused by melt pool surface fluctuations.Gray-level image threshold method is an intensity based method and is found difficult to deal with images with strong intensity-related uncertainties,such as changing brightness,localized fuzzy areas and speckle disturbances of a melt pool from a laser material processing.On the contrary,phase-based image processing considers the phase components of the Fourier transformed image and has the advantage to perform well over intensity when the boundary phase information is invariant to its intensity.Therefore,phase-based image processing has the potential to extract more accurate melt pool geometry of additive manufacturing process than intensity-based methods.The aim of this work is to investigate a phase-based method to capture the melt pool boundary during a laser additive manufacturing process.In the following sections,experimental details are firstly provided.Then the melt pool edge extraction using phase congruency with log-Gabor wavelets are introduced.Next,the performance of melt pool edge extraction using phase congruency is compared with fixed gray-level threshold method and Otsu threshold method,and the performance improvement of melt pool edge extraction with phase congruency over the other two methods are discussed.Melt pool geometry,as one of the most important built attributes that reflect the stability of an additive manufacturing process,has been widely used for process monitoring and process control.Currently,intensity-based gray-level image processing is generally used to extract the melt pool boundaries.However,selection of proper threshold is challenging due to the intensity-related noise and disturbance from blurred areas,flares and speckles.This thesis presents a phase congruency melt pool edge extraction approach that performs well for phase-invariant but intensity-variant image boundary extraction.The results show that the phase congruency edge extraction approach can not only robustly handle disturbances from blurred areas,flares,black areas and incandescent droplets,but also obtain more accurate melt pool geometries.A real-time melt pool monitoring system is developed to monitor a geometrically dependent and time varying melt pool evolution during an additive manufacturing process.