Analysis for real-time hybrid simulation in the presence of feedback sensor delay and data loss.

Abstract

Real-time hybrid simulation (RTHS) studies typically use wired sensors familiar to civil engineering, such as linear variable differential transformers (LVDT), load cells and strain gauges. However, test specimens with complex or difficult to instrument geometries may see the benefit from noncontact or wireless sensing. These methods cannot be applied without preliminary investigation, as they can introduce time delay and information loss into the sensor feedback loop. Delay and loss in the sensor feedback loop can cause stability and performance issues that need to be compensated. The Benchmark control problem for real-time hybrid simulation from Silva et. al served as a case study for this work. The Smith Predictor was presented and validated as a compensation method for the case of sensor delay. A key performance metric was the root-mean-square (RMS) between the RTHS structure's displacement and that of the reference structure for the three degrees-of-freedom (DOF) for the simplified system model. These performance metrics for the delayed system were improved by 20\% at least when the Smith Predictor was used. To compensate for signal loss, the Time Varying Kalman Estimator (TVKE) and a class of estimators known as Finite Loss History Estimators (FLHE) were tested using a Bernoulli process model. The same performance metrics were improved by 20-30\% at least between lossy cases with and without compensators. The FLHE with the shortest history length of $l = 1$ was determined to be sufficient for compensating information loss, as overall simulation performance with this estimator was within 5\% of its competitors and it is the most computationally efficient. Taken together, these results can be used to inform future work regarding noncontact or wireless sensors in the sensor feedback loop of RTHS's, with the end goal being to implement these sensors into a RTHS laboratory setting.