Tool Delivery Prediction through Adaptive Neuro-Fuzzy Inferencing

The development of a tool management system for reliable tool delivery prediction in a supply network for optimum tool inventory sizing based on reliable delivery forecasting of Cubic Boron Nitride (CBN) grinding wheels for nickel base alloy turbine blade fabrication is illustrated in this work. The basis for the development of the system is represented by the historical data on tool management including the chronological series of CBN grinding wheel shipment and delivery dates between one manufacturing company (client) and several tool manufacturers (suppliers) in a supply network. If historical data are highly variable, stochastic management methods, such as time series analysis, are either inapplicable or responsible for excessive inventory sizing. Alternative approaches are given by special analysis and modelling methodologies, such as those based on fuzzy logic, which deal with deterministic events but are also capable to take into account unpredictable factors for better results in prediction and forecast. In this paper, supplier-dependent dressing cycle time predictions for each external tool manufacturer in the supply network are obtained through a neuro-fuzzy approach whose structure is a 1st order Sugeno fuzzy model known as Adaptive Neuro-Fuzzy Inference System (ANFIS). The ANFIS predictions can be utilized by the customer to evaluate the supplier reliability in the delivery of CBN grinding wheels, which represents a critical decision parameter in the dressing order allocation procedure and a key reference factor for the implementation of flexible tool management strategies.