Learning control strategies for high-rate materials testing machines

Hydraulic high strain rate materials testing machines are required to track a user-defined velocity profile during tensile or compression tests in the face of sudden large impact forces. Due to delays and limited bandwidth of the actuation system, causal feedback/feedforward controllers fail to compensate for these disturbances. This paper presents more suitable non-causal learning control strategies, which anticipate the impact and take corrective action in advance. Two control strategies are discussed. The first comprises an iterative algorithm, which calculates a command signal correction by passing the velocity error observed in the previous test through an inverse model linearized around the target velocity. In the second approach, a detailed nonlinear inverse model is used to obtain a command signal from demand motion and force data. It is concluded that the first method is superior if two or more iterations can be performed.