Missing sensors restoration for system control and diagnostics

Error free measurements are an essential requirement for system monitoring, diagnosis, and control. Measurements can be corrupted or interrupted due to sensor failure, broken links, or bad communication. Control, monitoring and diagnostics cannot operate effectively under these conditions. State estimation has been used in the past for sensors restoration. However, it requires accurate, observable and error free detailed system model. These requirements are often unattainable in hardware settings. In this paper, we propose a completely different concept based on auto-encoding and intelligent system search algorithms. The proposed technique is not model based, is hardware realizable, and is rapid enough for fast action of low inertia electromechanical systems. Hardware experimental results show the effectiveness of this technique for on-line control.