A Data-Driven Fault Diagnosis Method using Modified Health Index and Deep Neural Networks of a Rolling Bearing

To improve fault diagnosis accuracy, a data-driven fault diagnosis model based on the adjustment Mahalanobis-Taguchi system (AMTS) was proposed. This model can analyze and identify the characteristics of vibration signals by using degradation monitoring as the classifier to capture and recognize the faults of product more accurately. To achieve this goal, we firstly used the modified ensemble empirical mode decomposition (MEEMD) scalar index to capture the bearing condition; then, by using the key intrinsic mode function (IMF) extracted by AMTS as the input of classifier, the optimized properties of bearing is decomposed and extracted effectively. Next, in order to improve the accuracy of the fault diagnosis we tested different modes; employing the modified health index (MHI), which is designed to overcome the shortcomings of the proposed health index as a classifier in single fault mode, and the deep neural networks (DNN) as a classifier in multi-fault mode. To evaluate the effectiveness of our model, the Case Western Reserve University (CWRU) bearing data were used for verification. Results indicated a strong robustness with 99.16% and 1.09s, 99.86% and 6.61s fault diagnosis accuracy in different data modes respectively. Furthermore, we argue that this data-driven fault diagnosis obviously lowers the maintenance cost of complex systems by significantly reducing the inspection frequency and improves future safety and reliability.

A Failure Analysis Method Based on the Combination of FTA and GGRA: A Case Study on an Engine

As product structure becomes more and more complex, the fault mode presents a diversified trend, and it is more difficult to determine the causes of system failure for a complex product. The main objective of this study is to provide an effective failure analysis method based on the combination of fault trees analysis (FTA) and generalized grey relation analysis (GGRA) for complex product. In this method, the product system failure is defined and the fault tree is constructed by FTA methodology firstly; and then GGRA is employed to identify the correlations between each fault mode and the system failure; finally, the main causes of system failure are identified and the corresponding measures can be made. A case study of a WD615 Steyr engine is conducted throughout the text to verify the validity of this method. The present study would help facilitate the failure and reliability analysis for complex product and benefit designers for the product improvement.

Experimental Validation of Model-Based Prognostics for Pneumatic Valves

Because valves control many critical operations, they are prime candidates for deployment of prognostic algorithms. But, similar to the situation with most other components, examples of failures experienced in the field are hard to come by. This lack of data impacts the ability to test and validate prognostic algorithms. A solution sometimes employed to overcome this shortcoming is to perform run-to-failure experiments in a lab. However, the mean time to failure of valves is typically very high (possibly lasting decades), preventing evaluation within a reasonable time frame. Therefore, a mechanism to observe development of fault signatures considerably faster is sought. Described here is a testbed that addresses these issues by allowing the physical injection of leakage faults (which are the most common fault mode) into pneumatic valves. What makes this testbed stand out is the ability to modulate the magnitude of the fault almost arbitrarily fast. With that, the performance of end-of-life estimation algorithms can be tested. Further, the testbed is mobile and can be connected to valvesin the field. This mobility helps to bring the overall process of prognostic algorithm development for this valve a step closer to validation. The paper illustrates the development of a model-based prognostic approach that uses data from the testbed for partial validation.

Controlling Tracking Performance for System Health Management - A Markov Decision Process Formulation

After an incipient fault mode has been detected a logical question to ask is: How long can the system continue to be operated before the incipient fault mode degrades to a failure condition? In many cases answering this question is complicated by the fact that further fault growth will depend on how the system is intended to be used in the future. The problem is then complicated even further when we consider that the future operation of a system may itself be conditioned on estimates of a system’s current health and on predictions of future fault evolution. This paper introduces a notationally convenient formulation of this problem as a Markov decision process. Prognostics-based fault management policies are then shown to be identified using standard Markov decision process optimization techniques. A case study example is analyzed, in which a discrete random walk is used to represent time-varying system loading demands. A comparison of fault management policies computed with and without future uncertainty is used to illustrate the limiting effects of model uncertainty on prognostics-informed fault management policies.