Intelligent diagnostics and prognostics for industrial machines using an optimization approach

The complexity and high-value of many industrial machines coupled with the ever-increasing demands on manufacturers to offer ensured operational availability means that effective ways of monitoring their health must be adopted. One way to achieve more accurate diagnostic capabilities is to use pattern matching and classification techniques which allow the identification of novel behavior. This paper proposes a new data discretization algorithm based on the Piecewise Aggregate Approximation (PAA) algorithm called Optimum-PAA, in order to facilitate more accurate matching of operational data and also allow identification of particular failure modes. The paper also demonstrates how the performance of the Symbolic Aggregate Approximation (SAX) pattern matching algorithm is increased through the use of the Optimum PAA parameters.