Comparing various friction models of hydraulic cylinders based on accuracy, parameter dependency, and efficiency for heavy-duty material handlers

Abstract

Friction has detrimental effects on mechanical systems, especially for those that require high accuracy and precision like hydraulic cylinders. As such, friction models have been created in order to analyse, calculate and predict friction. This thesis aims to recreate the Brown – McPhee (BMP) model, the LuGre model and the Modified LuGre (MLG) model in Simulink and compare their parameter dependency, accuracy and efficiency. Both simplified models and complete hydraulic systems are used for assessment. Due to the absence of physical experiments, the results and conclusions are only derived from behaviours of simulated systems. This is especially true in the accuracy tests where the BMP model is used as a baseline for relative difference to compare against the other models. It is revealed that the two LuGre models have smaller maximum friction at stiction phase, though they are also more vulnerable to static friction overshoots when undergoing rapid, large velocity changes. For parameter dependency data, it is found that during stiction, the BMP model’s most influential parameter is static friction value while it is bristle damping value for the LuGre and MLG models. During sliding friction, the models share a similar trend with Coulomb friction being the most influential parameter. In the efficiency tests, the BMP model is the most efficient as it requires the smallest execution time, is capable of computation using larger average step sizes, and is easy to implement. On the contrary, the MLG model has the largest execution time, requires smaller average step sizes and is the most complicated to be assembled. Regarding energy loss, the amount is approximately the same for all models, but MLG has the smallest loss. It is concluded that the BMP model is suitable for simple systems, the MLG model is suitable for complex, precise systems with computational resources while the LuGre model can act as middle ground between the other models.