With the recent advances in digital technologies, the subtractive manufacturing industry is striving for smart machine tools, capable of data-driven self-optimization. As a building block, this work proposes an approach for incorporating awareness regarding the material and its batch-specific characteristics for process optimization. The proposed smart manufacturing system utilizes cutting tool images for an initial condition assessment. Methods are proposed for the semantic segmentation of the defect classes encountered in tool condition monitoring, enabling a detailed analysis regarding their presence, location, and size. Furthermore, novel methods are proposed that support the image annotation process and the adaptation of existing training data to new scenes. During machining, internal control data is used for material batch identification. The high-frequency control data is preprocessed, error-compensated, and aggregated into features. Using a novelty detection algorithm, unknown batches are identified. Subsequently, a classification algorithm is used to classify known batches, whereas a clustering approach is used to analyze unknown batches. In a final step, historic process knowledge is used to compute optimized cutting parameters, thus enabling batch-adaptive machining. Furthermore, operational routines are proposed for the automated incorporation of material batches with novel behavior, continuous model improvement, and efficient adaptation to new machining scenarios.

This book is included in DOAB.