Fuqiang Sun

Utilizing Accelerated Degradation and Field Data for Life Prediction of Highly Reliable Products

For newly developed, highly reliable, and long-lifespan products, it is quite difficult to implement effective remaining useful life (RUL) prediction in the early usage under limited time cost. However, accelerated degradation testing (ADT) is generally used for lifetime evaluation for such products with harsher test conditions and shorter test time in the late research and development phase. Thus, in this paper, we propose a life prediction framework to integrate the information from ADT to conduct field RUL prediction for highly reliable products. Because ADT belongs to reliability testing used for inferring the population information from the selected test samples, we at first present the modified Wiener process (MWP) model. Different from traditional methods that embody both the random variability and unit-to-unit variability into the diffusion coefficient, the proposed method describes them separately in ADT analysis. Then, the MWP model from ADT is used as a prior for field RUL prediction of the target product during which the strong tracking filtering algorithm is introduced for updating the hidden state and computing the RUL prediction results when the new monitoring data are available. Because of the complexity of the MWP model, the Markov chain Monte Carlo method is provided to estimate the unknown parameters. Finally, the simulation study and the light-emitting diode application verify the effectiveness of the proposed framework that can achieve reasonable life prediction results for highly reliable products for both linear and nonlinear scenarios. Copyright

An Analytical Model for Fatigue Crack Propagation Prediction with Overload Effect

In this paper a theoretical model was developed to predict the fatigue crack growth behavior under the constant amplitude loading with single overload. In the proposed model, crack growth retardation was accounted for by using crack closure and plastic zone. The virtual crack annealing model modified by Bauschinger effect was used to calculate the crack closure level in the outside of retardation effect region. And the Dugdale plastic zone model was employed to estimate the size of retardation effect region. A sophisticated equation was developed to calculate the crack closure variation during the retardation area. Model validation was performed in D16 aluminum alloy and 350WT steel specimens subjected to constant amplitude load with single or multiple overloads. The predictions of the proposed model were contrasted with experimental data, and fairly good agreements were observed.

A multi-axis vibration fixture based on electromagnetic shaker

This paper introduces a kind of vibration fixture which can support products undergo vibration simultaneously along three mutually orthogonal axes on the electromagnetic shaker, and the ratio of the vibration values along three product installation directions is coincident with the theoretical set value. Moreover, the vibration feature and design scheme of this multi-axis vibration fixture are discussed. It is found that using this multi-axis vibration fixture on the electromagnetic shaker can avoid the problems currently existing in the vibration testing using the pneumatic hammer shaker and the electromagnetic shaker, such as random vibration frequency spectrum of the pneumatic hammer shaker is uncontrollable and the electromagnetic shaker can only perform uniaxial vibration testing, etc. Finally, the vibration feature of the multi-axis vibration fixture is verified by a testing. (Abstract)

A General Accelerated Degradation Model Based on the Wiener Process

Accelerated degradation testing (ADT) is an efficient tool to conduct material service reliability and safety evaluations by analyzing performance degradation data. Traditional stochastic process models are mainly for linear or linearization degradation paths. However, those methods are not applicable for the situations where the degradation processes cannot be linearized. Hence, in this paper, a general ADT model based on the Wiener process is proposed to solve the problem for accelerated degradation data analysis. The general model can consider the unit-to-unit variation and temporal variation of the degradation process, and is suitable for both linear and nonlinear ADT analyses with single or multiple acceleration variables. The statistical inference is given to estimate the unknown parameters in both constant stress and step stress ADT. The simulation example and two real applications demonstrate that the proposed method can yield reliable lifetime evaluation results compared with the existing linear and time-scale transformation Wiener processes in both linear and nonlinear ADT analyses.

Modified Norris–Landzberg Model and Optimum Design of Temperature Cycling Alt

Accelerated life testing (ALT) is an effective way to assess the lifetime of a product. Due to the complex nature of its testing profile, it is difficult to carry out temperature cycling ALT. This paper establishes a modified Norris–Landzberg model as acceleration model, and proposes the optimum design method of temperature cycling ALT. First, the FEA method is used to study the influence of temperature cycling profile parameters on the thermal fatigue life of 63Sn–37Pb solder joints. Then, a modified Norris–Landzberg model is proposed by introducing ramp time and dwell time with an added weight value. Finally, the temperature cycling ALT is regarded as a special multi-stress ALT to study its optimum design method. The uniform design theory is used to determine the combined mode. The optimum model is established with the objective of minimizing the asymptotic variance of the estimation of median lifetime under normal use conditions, and the simulation example shows the workability of the proposed method.

Life prediction of jet engines based on LSTM-recurrent neural networks

The issue of remaining useful life (RUL) prediction has already become a quite interesting topic in industrial product. The data driven RUL prediction has been applied to the current research by taking advantage of a long-short term memory (LSTM)-recurrent neural network (RNN) approach. This means that even in a specified long-short term memory bound and limited available data sets, the RUL predictions can also improve the equipment capacity. By collecting the sensor parameters from National Aeronautics and Space Administration (NASA) jet engines, the capability of this approach has been demonstrated. An appropriate selection, that is suitable for the variable measurement, has been used to feed LSTM-RNNs with the most useful RUL labels. The analysis results illustrate that, compared with traditional statistical model, the development prediction approach is likely to provide an accurate prediction for RUL equipment that can be meaningful to maintenance schemes.

The temperature fluctuation modeling and compensation for the degradation data of super-luminescent diode

The environmental factor (such as temperature, etc.) is an important error sources of product performance degradation data. In practice, the raw degradation data usually contain some noise terms caused by the fluctuation of environmental factor. Therefore, it is necessary to extract the real degradation trend before degradation modeling and life prediction. In this research, the Grey relational analysis method is utilized to quantitatively describe the correlational relationship between temperature fluctuation and raw degradation data. Then a novel temperature compensation model based on least squares support vector machine (LS-SVM) is proposed, which can be used to compensate the influence of environmental fluctuation on degradation data. An engineering case study on the degradation data of a super-luminescent diode (SLD) is employed to verify the effectiveness of the proposed method.

Remaining Useful Life Prediction for a Machine With Multiple Dependent Features Based on Bayesian Dynamic Linear Model and Copulas

Degradation modeling and remaining useful life (RUL) prediction for products with multiple degradation features are hot topics in the prognostic and health management. The key to this problem is to describe the dependence among multiple degradation features effectively. In this paper, a multivariate degradation modeling approach based on the Bayesian dynamic linear model (BDLM) is proposed to calculate the RULs of degradation features, and the Copula function is employed to capture the dependence among RUL distributions. A combined BDLM is used to establish the multivariate degradation model, which includes two typical BDLMs, namely, the linear growth model and seasonal factors model. After the model parameters get calibrated by the maximum likelihood estimation, the model can predict the degradation process of features. Once the failure thresholds are given, the probability density function and cumulative distribution function (CDF) of RUL for each degradation feature can be obtained. Since these RUL distributions are not independent of each other, the Copula function is adopted herein to couple the CDFs. Finally, some practical testing data of a microwave component, which has two degradation features, are utilized to validate our proposed method. This paper provides a new idea for the multivariate degradation modeling and RUL prediction.

A degredation interval prediction method based on RBF neural network

In the area of reliability, remaining useful lifetime (RUL) prediction can help people establish reasonable maintenance strategies and then implement maintenance activities at a right time. In this paper, RBF neural network approach is applied in the degradation prediction process of a certain microwave component. A degradation model that describes how a certain degradation parameter changes over time is established and then the performance degradation trend can be obtained based on this model. And then a confidence interval prediction can be obtained based on traditional probability theory, which proves that the results have reached a high confidence level. Finally, the BP neural network approach is introduced as a comparison, and results indicate that the proposed method has higher precision and stability.

Life and Reliability Prediction of the Multi-Stress Accelerated Life Testing Based on Grey Support Vector Machines

There are many difficulties in statistical analysis of multi-stress accelerated life testing, such as establishing the accelerated model and solving pluralism likelihood equations. With a focus on these difficulties, the Grey-SVM based life and reliability prediction method for multi-stress accelerated life testing is proposed, with the accelerated stress level and the reliability as SVM inputs, and the corresponding Grey AGO processing failure data as outputs. Simulation and case study shows that the method has high prediction accuracy and with less amount of training samples than neural network.