Model-based tool condition prognosis using power consumption and scarce surface roughness measurements

In machining processes, underusing and overusing cutting tools directly affect part quality, entailing economic and environmental impacts. In this paper, we propose and compare different strategies for tool replacement before processed parts exceed surface roughness specifications without underusing the tool. The proposed strategies are based on an online part quality monitoring system and apply a model-based algorithm that updates their parameters using adaptive recursive least squares (ARLS) over polynomial models whose generalization capabilities have been validated after generating a dataset using theoretical models from the bibliography. These strategies assume that there is a continuous measurement of power consumption and a periodic measurement of surface roughness from the quality department (scarce measurements). The proposed strategies are compared with other straightforward tool replacement strategies in terms of required previous experimentation, algorithm simplicity and self-adaptability to disturbances (such as changes in machining conditions). Furthermore, the cost of each strategy is analyzed for a given benchmark and with a given batch size in terms of needed tools, consumed energy and parts out of specifications (i.e., rejected). Among the analyzed strategies, the proposed model-based algorithm that detects in real-time the optimal instant for tool change presents the best results.