Multi-level interpretable logic tree analysis: A data-driven approach for hierarchical causality analysis

Abstract

This paper presents a data-driven approach for a hierarchical causality analysis of faults in a complex system, named a multi-level interpretable logic tree (MILTA). From a representative faults dataset, this approach constructs dependent trees that explain the relation structure between the root-causes, intermediate causes and faults with the minimum expert involvement. The MILTA model combines the discovered knowledge in dataset (KDD) in the form of feasible solutions and the fault tree analysis (FTA), level after level, as long as the root-causes are not completely uncovered. A burn-and-build algorithm is developed to maximize the representability of the feasible solutions with a minimum number of patterns. Using Bayes’ theorem, the hierarchical causality between the root-causes and the fault is captured through different causality rules that quantify the effects of the root-causes on the fault occurrence. An actuator system dataset that consists of complex fault and normal operation states is used as an illustrative example. The MILTA model finds the same documented root-cause and uncovers other root-causes with higher accuracy.