A ROBUST CONTROL CONCEPT FOR HYDRAULIC DRIVES BASED ON SECOND ORDER SLIDING MODE DISTURBANCE COMPENSATION

The application of sliding mode algorithms for control of hydraulic drives has gained increasing interest in recent years due to algorithm simplicity, low number of parameters and possible excellent control performance. Both application of first and higher order sliding mode control algorithms in hydraulic drive controls have been presented in various literature, demonstrating the possible successful application of these. However; a major drawback is the presence of e.g. valve dynamics which often necessitates the usage of continuous approximations of discontinuities in order to avoid control chattering, which on the other hand compromises the robustness properties. This may also be the case when discontinuities are only present in the control derivative. This fact suggests that sliding mode algorithms may be more appropriate for assisting the control, i.e. for state observation, disturbance observer based control etc., and several examples of such approaches have been presented in literature. The latter case appear especially interesting as a sliding mode actually takes place, but only the low-pass filtered sliding mode algorithm output is used in the actual control input. However, the successful implementation relies heavily on the low-pass filter design where the drive dynamics, sample rate etc. play a significant role. In this paper the utilization of the super twisting algorithm for disturbance compensation is considered. The fact that the discontinuity here is nested in an integral causes less restrictions on the used low-pass filter, enabling the possibility for a wide range of usage when the sampling frequency is in the range of industrial grade control hardware. The proposed control structure is designed and considered in relation to a valve driven hydraulic winch drive, and results demonstrate that excellent and high precision control may be achieved in the presence of large and highly varying external loads.