The success of unattended manufacturing depends largely on control mechanisms that monitor the machining state and take actions to rectify unsatisfactory performance. Direct sensing methods like quality inspection lack on-line capability, whereas indirect methods using sensors can be thwarted by noise and changes in operating conditions. While knowledge about these changes exists, it does not generally correspond with an available sensor. Two different techniques are applied to the problem of integrating data from multiple sensors in the manufacturing environment: one featuring the integration of fuzzy logic and neural networks, and one using a probabilistic neural network. These techniques are applied to monitor and diagnose tool wear in unattended milling machines - an application with implications toward extension to other manufacturing machines.