Data driven prognosis approach for safety critical systems

Safety critical systems are being developed to improve the performance and cost effectiveness. The safety critical system are used in various domain such as aerospace domain, military, defense etc. In an aerospace domain there are many parameters affects the system environmental conditions, or hazards which cause many faults in the system which leads to failure. It is necessary to know before the system fails, so that necessary remedies can take to prevent the failure. The tool/software is needed to monitor the health management of safety critical systems. In this paper a prognostic technique is being used to mitigate the system failure. There are many techniques for the prognosis such as data driven technique, model based technique, and hybrid technique. This paper proposes implementation of the artificial neural network [ANN] based prognosis illustrates the use of data driven technique. The novelty of the proposed algorithm is that it uses formal techniques to develop a robust & reliable prognostics algorithm. The approach developed will be demonstrated for gyro sensor a critical component in the aerospace domain. The ANN can train and classify real data from the gyro sensors, and it is implemented using high level interpreted language GNU-Octave. The cost function/error function is calculated for the trained ANN data and it is being observed that the values are converging to the minimum value. At last the system is classified as healthy, partially healthy, and unhealthy state of the system.