When a human controls a manufacturing process he or she uses multiple senses to monitor the process. Similarly, one can consider a control approach where measurements of process variables are performed by several sensing devices which in turn feed their signals into process models. Each of these models contains mathematical expressions based on the physics of the process which relate the sensor signals to process state variables. The information provided by the process models should be synthesized in order to determine the best estimates for the state variables. In this paper two basic approaches to the synthesis of multiple sensor information are considered and compared. The first approach is to synthesize the state variable estimates determined by the different sensors and corresponding process models through a mechanism based on training such as a neural network. The second approach utilizes statistical criteria to estimate the best synthesized state variable estimate from the state variable estimates provided by the process models. As a “test bed” for studying the effectiveness of the above sensor synthesis approaches turning has been considered. The approaches are evaluated and compared for providing estimates of the state variable tool wear based on multiple sensor information. The robustness of each scheme with respect to noisy and inaccurate sensor information is investigated.