Root cause analysis (RCA) is a method of problem solving used for identifying the root causes of faults or problems. RCA is widely used in industrial process control, IT operations, health industry and accident analysis in aviation, nuclear plants and rail transport. Without delving in the intricate details of specific problems, several general conditions can make RCA more difficult than it may appear at first sight. It is generally not possible, in practice, to monitor everything manually. There may be more than one root causes for a given problem and causal graphs often have many levels which makes the causal graph very difficult to establish and query.
We developed an AI based solution of finding root cause and most influential path of the fault in industrial plants by learning the structure of causal Bayesian network from plant-data/sensor measurements and then using causal inference for finding causal effects and path specific effects. A flexible and modular solution is designed for online monitoring of plants, predictive maintenance, finding the effects of changing the values of control nodes on other nodes in the plant. Solution is flexible enough to handle latent variables and works with both categorical and numeric variables/nodes.