No abstract

No abstract

In this paper, a statistical prognostic method to predict the remaining useful life (RUL) of individual units based on noisy condition monitoring signals is proposed. The prediction accuracy of existing data-driven prognostic methods depends on the capability of accurately modeling the evolution of condition monitoring (CM) signals. Therefore, it is inevitable that the RUL prediction accuracy depends on the amount of random noise in CM signals. When signals are contaminated by a large amount of random noise, RUL prediction even becomes infeasible in some cases. To mitigate this issue, a robust RUL prediction method based on constrained Kalman filter is proposed. The proposed method models the CM signals subject to a set of inequality constraints so that satisfactory prediction accuracy can be achieved regardless of the noise level of signal evolution. The advantageous features of the proposed RUL prediction method is demonstrated by both numerical study and case study with real world data from automotive lead-acid batteries.

One of the common ways to perform data-driven fault diagnosis is to employ statistical models, which can classify the data into nominal (healthy) and a fault class or distinguish among different fault classes. The former is termed fault (anomaly) detection, and the latter is termed fault isolation (classification, diagnosis). Traditionally, statistical classifiers are trained using data from faulty and nominal behaviors in a batch mode. However, it is difficult to anticipate, a priori, all the possible ways in which failures can occur, especially when a new vehicle model is introduced. Therefore, it is imperative that diagnostic algorithms adapt to new cases on an ongoing basis. In this paper, a unified methodology to incrementally learn new information from evolving databases is presented. The performance of adaptive (or incremental learning) classification techniques is discussed when: 1) the new data has the same fault classes and same features and 2) the new data has new fault classes, but with the same set of observed features. The proposed methodology is demonstrated on data sets derived from an automotive electronic throttle control subsystem.