Integrated Prognostic Model for RUL Estimation using Threshold Optimization

Abstract

The capability of prognostic models to estimate loss of machines functionality, enables plant operators to manage the system maintenance and related logistics tasks (spare part management) effectively. Prognostic models will be a key feature and therefore connect the operation of machines and systems in higher modes of automation and factory digitalization, denoted as Industry 4.0 (Germany), Industrial Internet (US), and China 2025 (China) focusing to machines life cycle with respect to reliable and continuous operation and functionality. Prognostic models allow to estimate reliability characteristics mainly related to lifetime and reliability’s characteristic (hazard rate, availability). Using the models during operation, operational cost and in the best case due to connected maintenance strategies unscheduled machine downtimes can be reduced. Through appropriate control strategies, it is possible to preserve the service lifetime based on the information of damage accumulation in unpredicted circumstances. Even with inadequate informations extracted from monitoring data, prognostic schemes allow to predict upcoming physical characteristics that permits a higher level of condition-based system maintenance. The optimization of integrated lifetime model becomes vital for accurate estimation of machine’s state and therefore the Remaining Useful Lifetime (RUL). In this work, a previously developed, novel state-machine model combined with parametric approaches is extended. The approach can be characterized as parameterized state machine model. During training, experimental data including stochastically effects stress-related data are combined with the moment in time of functional loss; related parameters, and thresholds are combined to define a model by optimization. The threshold optimization will identify the best optimal solution. The application (test) for unknown system data demonstrate that these new kind of prognostic model is able to estimate the RUL with high accuracy. In this paper, the optimization of the model containing the structure of state machine as well as related parameters are discussed. Experimental results demonstrate the effectiveness of the improvement.