A data-driven simulation and Gaussian process regression model for hydraulic press condition diagnosis
Denis Jankovič, Marko Šimić, Niko Herakovič
Abstract
Improving overall performance and increasing operational reliability are currently among the leading research topics in the field of hydraulic systems. In recent years, the use of artificial intelligence-based modeling and design techniques has developed rapidly to account for the nonlinear properties of Gaussian systems and to predict fault reasoning in hydraulic systems. In this study, feature acquisition and selection are proposed to prepare input data for a simulation-based learning approach. In addition, a cause-and-effect analysis is performed by considering various what-if scenarios as external disturbances that affect the response of the hydraulic press. While the objective of the sheet metal bending cycle and a pulley system is to initiate a load on the hydraulic press, an intelligent sensing system is used to observe the behavior of the hydraulic press during the phases of sheet metal bending cycle, i.e., the forming, leveling, and movement. In addition, the Gaussian process regression method is used to build data-driven prediction models with different predictors that contribute significantly to improving predictive accuracy. The condition diagnosis indicates the accurate performance of predictive models observing the coefficient of determination R2 at 0.998 for the bending phase, 0.962 for the leveling phase, and 0.999 for the movement phase. Although the approximation of the simulation model is efficient, it is found that certain features are reasonably well approximated with regard to the forming phases.