Model-in-the-Loop (MiL) testing is a method in which the test object is split into a physical part and a simulated part, and these are connected with interfaces to form a combined physical-numerical system. It is introduced to combine the advantages of physical test and computer simulation: the part of the system which is difficult to implement physically can be put into the numerical subsystem to reduce the cost and complexity of the physical test, and the key components with unknown characteristics or with some characteristics which are difficult to model can form the physical subsystem. The simulated part also provides the flexibility to change the parameters during the test.In this thesis, the structure and the characteristics of MiL systems are analysed. Detailed results are given using two example systems: a single mass-spring-damper MiL system, and a two Degree of Freedom (DOF) mass-spring-damper MiL system. The systems are defined, and a procedure for stability analysis is given. The influence of the actuator dynamics and the measurement noise introduced by the sensors is discussed. To compensate for the actuator dynamics, compensators are introduced to the MiL system. It is shown with simulation results that, when a compensator based on an inverse of the actuator dynamics is added to the MiL system, the high frequency measurement noise may be greatly amplified in the compensated signal, and therefore signal saturation may occur which leads to unacceptable testing results.To design a compensator which can effectively compensate for the actuator dynamics, while reducing the tendency of signal saturation in the compensated actuator control signal at the same time, H∞ optimization is applied. A general model is composed for the H∞ optimization, where the target testing result is compared with that of an ideal reference model, and the error between them is minimized via H∞ loop shaping. The principle of H∞ loop shaping is presented in the thesis, and its use as a general MiL optimization procedure is proposed. The optimization method is verified with the example one and two DOF mass-spring-damper MiL systems. The simulation results show that, for both of the examples, the H∞ optimized compensator can compensate for the actuator dynamics accurately, and attenuate the response excited by the measurement noise in the compensated signal effectively. The balance between accuracy and high frequency noise attenuation can be adjusted by the weighting functions.The effectiveness of the H∞ optimized compensator is then verified with experimental results. A two-axis robotic arm based on a limb of the Italian Institute of Technology HyQ robot was used for the experiment. The H∞ optimized compensator is compared with various alternative compensators, and the H∞ optimized compensator show its advantages in terms of an appropriate balance between accuracy and saturation rejection.Lastly, a performance envelope analysis is introduced to give a guide to choosing suitable hydraulic actuators and valves for a specific MiL test based on the actuator performance required to give desired test accuracy. Although the H∞ optimized compensator can broaden the usable frequency range of the valve/actuator system, and will provide a larger margin for control signal saturation, an effective test system is only achievable if an actuation system of adequate performance is chosen.
|Date of Award||6 Dec 2017|
|Supervisor||Andrew Plummer (Supervisor)|