Remaining useful life prediction of rolling element bearings using artificial intelligence

Abstract

To remain competitive, manufacturers of production plants and machinery are required to continuously keep their supply chain management smooth and running by newlinetending to the availability, reliability and security of their traditional product-focused newlinebusinesses while reducing their maintenance cost. In such a scenario, prognostics and newlinehealth management (PHM) plays a major role. In this domain the condition-based newlinemaintenance (CBM) and predictive maintenance (PM) are efficient maintenance newlinestrategies that allow optimizing the maintenance through estimation of the time to newlinefailure (ETTF). Indeed, contrary to a traditional corrective maintenance where the newlineinterventions are done after the occurrence of the failure, in a CBM (or a PM), the newlineinterventions are carried out according to the observed or estimated health condition newlineof the equipment. Generally, a CBM system is seen as the integration of seven layers: newlinesensors, signal processing, condition monitoring (or fault detection), health assessment (or fault diagnostics), prognostics, decision support and finally presentation newlinelayers. Among these activities, research on failure prognostics has gained much newlineindustrial interest. This has in turn led to numerous methods, tools and applications during the last decade. According to the reported literature, failure prognostic newlinemethods can be classified into three main approaches: model-based, data-driven newlineand hybrid approaches. Model-based prognostic approach relies on the use of an newlineanalytical model (set of algebraic or differential equations) to represent the behavior newlineof the system including its degradation. The advantage of this approach is that it newlineprovides precise results. However, its drawback dwells in the fact that real systems newlineare often nonlinear and the degradation mechanisms are generally stochastic and difficult to represent in the form of analytical models