Application of multi sensor data fusion based on Principal Component Analysis and Artificial Neural Network for machine tool thermal monitoring

Due to the various heat sources on a machine tool, there exists a complex temperature distribution across its structure. This causes an inherent thermal hysteresis which is undesirable as it affects the systematic tool –to-workpiece positioning capability. To monitor this, two physical quantities (temperature and strain) are measured at multiple locations. This article is concerned with the use of Principal Component Analysis (PCA) and Artificial Neural Networks (ANN) to fuse this potentially large amount of data from multiple sources. PCA reduces the dimensionality of the data and thus reduces training time for the ANN which is being used for thermal modelling. This paper shows the effect of different levels of data compression and the application of rate of change of sensor values to reduce the effect of system hysteresis. This methodology has been successfully applied to the ram of a 5-axis gantry machine with 90 % correlation to the measured displacement.