A Trade-Off Analysis between Sensor Quality and Data Intervals for Prognostics Performance.

In safety-critical systems such as industrial plants or aircraft, failure occurs inevitably during operation, and it is important to prevent it in order to maintain high availability. To reduce this risk, a lot of efforts are directed from developing sensing technologies to failure prognosis algorithms to enable predictive maintenance. The success of effective and reliable predictive maintenance not only relies on robust prognosis algorithms but also on the selection of sensors or data acquisition strategy. However, there are not many in-depth studies on a trade-off between sensor quality and data storage in the view of prognosis performance. The information about (1) how often data should be measured and (2) how good sensor quality should be for reliable failure prediction can be highly impactful for practitioners. In this paper, the authors evaluate the efficacy of the two factors in terms of remaining useful life (RUL) prediction accuracy and its uncertainty. In addition, since knowing true degradation information is almost impossible in practice, the authors validated the use of the prognosis metric without requiring the true degradation information. A numerical case study is conducted to identify the relationship between sensor quality and data storage. Then, real bearing run-to-failure (RTF) datasets acquired from accelerometer (contact type) and microphone (non-contact type) sensors are evaluated based on the prognosis performance metric and compared in terms of the sensors' cost-effectiveness for predictive maintenance.

Challenges and Opportunities of System-Level Prognostics.

Prognostics and health management (PHM) has become an essential function for safe system operation and scheduling economic maintenance. To date, there has been much research and publications on component-level prognostics. In practice, however, most industrial systems consist of multiple components that are interlinked. This paper aims to provide a review of approaches for system-level prognostics. To achieve this goal, the approaches are grouped into four categories: health index-based, component RUL-based, influenced component-based, and multiple failure mode-based prognostics. Issues of each approach are presented in terms of the target systems and employed algorithms. Two examples of PHM datasets are used to demonstrate how the system-level prognostics should be conducted. Challenges for practical system-level prognostics are also addressed.

Information Value-Based Fault Diagnosis of Train Door System under Multiple Operating Conditions.

While there are many data-driven diagnosis algorithms for fault isolation of complex systems, a new challenge arises in the case of multiple operating regimes. In this case, the diagnosis is usually carried out for each regime for better accuracy. However, the problem is that different results can be derived from each regime and they can conflict with each other, which may invalidate the performance of fault diagnosis. To address this challenge, a methodology for selecting the most reliable one among the different diagnostic results is proposed, which combines the Bayesian network (BN) and the information value (IV). The BN is trained for each regime and a conditional probability table is obtained for probabilistic fault diagnosis. The IV is then employed to evaluate the value of several diagnostic results. The proposed approach is applied to the fault diagnosis of a train door system and its effectiveness is proven.

A Comparative Study of Fault Diagnosis for Train Door System: Traditional versus Deep Learning Approaches.

A fault diagnosis of a train door system is carried out using the motor current signal that operates the door. A test rig is prepared, in which various fault modes are examined by applying extreme conditions, as well as the natural and artificial wears of critical components. Two approaches are undertaken toward the fault classification for comparative purposes: one is the traditional feature-based method that requires several steps for the processing features such as signal segmentation, the extraction of time-domain features, selection by Fisher's discrimination, and K-nearest neighbor. The other is the deep learning approach by employing the convolutional neural network (CNN) to skip the hand-crafted features extraction process. In the traditional approach, good accuracy is found only after the current signal is segmented into the three velocity regimes, which enhances the discrimination capability. In the CNN, superior accuracy is obtained even by the original raw signal, which is more convenient in terms of implementation. However, in view of practical applications, the traditional approach is more useful in that the features processing can be easily applied to assess the health state of each fault and monitor the progression over time in the real operation, which is not enabled by the deep learning approach.