Abstract: This paper presents a 10 degree-of-freedom sensor system with embedded computing power and wireless interface. The motivation is to provide a high- performance sensor system for condition monitoring, which requires minimal effort for retrofitting in existing plant and machinery. The embedded computing power enables on board calibration of the sensor module. A computational method is presented for the independent identification of the different coordinate systems, which are implicit in all real application, i.e. sensor axis to sensor housing, sensor housing to machine in manufacturing environment and the machine to the local earth coordinate system during operation in the field. The embedded computational power is used to perform geometric elliptical approximations for vibrational orbits. This eliminates the possible Gibbs error associated with elliptical Fourier descriptors. Additionally, a new statistical approach is presented for identifying periodic signals in a highly perturbed environment. Results are presented from the monitoring of large vibrating sieves these results verify the correct functionality of the complete system and demonstrate the advantage of the new methods.

Keywords: Multi sensor system, Uncertainty analysis, Probability distribution, Condition monitoring, Retrofitting.
.__________________________________________________________________________________________