Work-in-Progress: A Probabilistic Approach to Discover Optimal Remedy Scheme for Industrial Diagnosis System

This paper presents an approach on cause and remedy diagnosis of system failures for general industrial purposes. We identify symptoms of failure in the system (which are captured using sensor data) and predict faulty components associated with the symptoms. The probabilistic engine computes the cost of remedy with and without testing of components and arrives at an optimal remedy scheme, the order in which components are to be fixed. After each component is remedied, the system is tested for its condition until it attains the working state. An application of this approach is explained with car starting mechanism. Our approach uses Bayesian Network as the probabilistic engine to arrive at optimal remedy scheme.