Data driven prognostics

Abstract

Condition monitoring and predictive analytics are well established in the field of industrial asset management. Diagnostic solutions, however, are naturally limited in their prognostic horizon: They provide thorough technical insights, but do not claim explicit, objective foresight. For prognostic purposes, most operators still rely on subjective expert gut feel. New prognostic technologies that simultaneously utilize continuous and periodical data streams work with different sets of data analysis methodologies. Based on unique stochastic algorithms, Prognostics solutions address this when question (when will the anomaly turn into a malfunction) and therewith enable the shift from time-based to condition-based maintenance. In contrast to established methods, Prognostics is non-parametric and not based on physical modelling; it also does not rely on fleet-Averages. No failure history is required such that also malfunctions that never occurred at that asset can be prognosticated. Utilizing condition and process data histories - including vibration, temperature, pressure, flow, etc. - Prognostics solutions generate and periodically update prognostic reports forecasting malfunction risk and remaining useful life. These prognostic reports support equipment operators with their decisions on optimal asset deployment, maintenance planning, and life cycle management. To illustrate the benefits of prognostics, detailed examples of prognostic use cases in different industries will be presented.