Software fault detection in control systems

Fault detection in the software of control systems is a difficult task. In largely interconnected systems, not only the individual performance of one channel, but also its interaction with other components, must be considered. In this conceptual paper, we outline a newmaintenance concept for the detection of software faults in control systems. The concept includes two approaches, morning gymnastics test and envelope analysis. The morning gymnastics test generates data for a baseline of the current operational abilities in contrast to the specified abilities and should be applied when feasible in continuous production systems. The test integrates historical and new sets of data to track degradation trends. Envelope analysis can be performed to detect operational anomalies and is based on subsequent deep analysis to distinguish software and hardware faults from each other. By using the envelope analysis it is possible to identify failures and disturbances affecting the control system. Thus, the proposed maintenance concept may facilitate detection and identification of potential failures in critical automated system.