Overinstrumented Systems

Advances in sensors and instrumentation have enabled the measurement of variables once thought to be inaccessible. Moreover, sensors have such high reliability that one can safely assume they will operate correctly. A case in point is the instrumentation in commercial aircraft with multiple built-in redundancy. We consider that it has become necessary to revisit the need for state observers. We submit that there is a significant number of dynamic systems where the complete state can be considered available. This class of systems we call overinstrumented and are the subject of our study. We show that it is possible to develop simple identification algorithms that permit speedy determination of system parameters. Moreover, the technique can be easily extended to include systems whose parameters can vary abruptly. This property in turn can be very important in the detection and identification of faults and in the development of adaptive and fault tolerant control algorithms. The paper develops the concept of observation model for the class of overinstrumented system. This model is linear in the system parameters and depends only on available measurements. The paper analyzes the observation model and establishes conditions for the model to be solvable for the system parameters. In particular, the paper shows that controllability is necessary. The paper analyzes the case study of a suspension system using the well known "quarter car model" on a Simulink implementation. Simulation results show that the technique permits identification of multiple changes in the value of the shock absorber constant.