Computer vision-assisted analysis of Ni-electroplated diamond wire wear and its effect on polycrystalline silicon sawn surface

Silicon plays a crucial role in the semiconductor industry, since it is an essential material for sensors and microprocessors, as well as in renewable energy technologies such as photovoltaic cells. However, the machining poses challenges due to its inherent brittleness and hardness, resulting in a high production cost. The multi-wire sawing is the main process used in the production chain, which is a precision abrasive machining that allows obtaining hundreds of wafers from a single silicon ingot. Nevertheless, the diamond wire undergoes intense wear, which is concerning due to it having only an abrasive monolayer. This study aims to analyze the wear of diamond wires used for slicing of polycrystalline silicon through a simplified computer vision method. Experiments were conducted on a continuous diamond wire sawing machine under constant cutting conditions, varying the slicing times in 7.5, 15, and 30 min. An algorithm was developed in Python to identify the wear of diamond grits by measuring the outer diameter of the diamond wires using scanning electron microscope images. Sliced surfaces were analyzed in terms of texture, morphology and surface roughness Sa and Sq to verify the wear effect on them. The results indicate that increased slicing time leads to greater wear on the diamond wire, as evidenced by reduction in the protrusion heights of grits and outer diameter of the diamond wire from 350.0 to 329.4 μm. Wear mechanisms include Ni-layer deformation and removal, exposed grits and abrasive wear. Once the grits have undergone rounding and flattening, there was ductile material removal. The brittle-to-ductile transition occurred when increasing the slicing time and, consequently, wearing the diamond wire which reduced surface roughness of the sliced poly-Si. Three main wear stages were identified: (i) initial Ni-layer wear; (ii) grits rounding and flattening which led to improved surface quality; and (iii) stable abrasive wear. The simplified computer vision algorithm demonstrated to be efficient for evaluating the wear of the diamond wires.