A multi sensor approach to in-process monitoring of surface roughness

Abstract

Presently single sensors are used for monitoring surface roughness. Different sensors sense roughness differently and hence there is an advantage in using multi sensing, to get a better evaluation and understanding of the roughness. This paper deals with a surface roughness assessment model using multiple sensors which can be used for surfaces produced by various manufacturing operations. An Artificial Neural Network is used to fuse three sensor signals. Here the Neural Network is trained to associate an input vector which consists of readings from three different sensors with an output vector from a standard stylus instrument. During operation Neural Network is fed with the sensor RMS values and it fuses this information and provides an estimated value of the parameters based on the model already established during training. The estimation was found to be reliable and can be extended to in-process monitoring applications also. Sensor selection sensor fusion training of the Network in-process monitoring and the results obtained are discussed in this paper.