Tongmin Jiang

A Bayesian reliability evaluation method with integrated accelerated degradation testing and field information

Accelerated degradation testing (ADT) is a common approach in reliability prediction, especially for products with high reliability. However, oftentimes the laboratory condition of ADT is different from the field condition; thus, to predict field failure, one need to calibrate the prediction made by using ADT data. In this paper a Bayesian evaluation method is proposed to integrate the ADT data from laboratory with the failure data from field. Calibration factors are introduced to calibrate the difference between the lab and the field conditions so as to predict a products actual field reliability more accurately. The information fusion and statistical inference procedure are carried out through a Bayesian approach and Markov chain Monte Carlo methods. The proposed method is demonstrated by two examples and the sensitivity analysis to prior distribution assumption.

Optimal design for step-stress accelerated degradation testing with competing failure modes

With the development of engineering and science technology, many products are able to operate for a long period of time before failure. Accelerated life testing (ALT) has been proposed to evaluate reliability and life of these products, but ALT may involve the collection of only a few failures (time-to-failure data). To overcome this problem, accelerated degradation testing (ADT) is presented. Competing failure mechanisms happen in many electronic devices. However, as for designing an efficient ADT experiment, competing failure mechanisms were seldom discussed. In this study, we first use drift Brownian motion to model a typical step-stress ADT (SSADT) problem. Then, according to competing failure rule, we established reliability model of the product. Next, under the constraint that the total experimental cost does not exceed a predetermined budget, our objective is to minimize the asymptotic variance of the estimated hundred percentile of the competing reliability model of product. This optimal testing plan gives the optimal number of test units and inspection times at each stress level. Finally, we analyze the different optimal plans with different budgets, different levels of stress and different stress steps. Based on these analyses, we propose the guideline of stress loading principles of SSADT.

Step-Stress ADT data estimation based on time series method

For long lifetime and high reliability products, it is difficult to obtain failure time data in a short time period. Hence, Accelerated Degradation Testing (ADT) is presented to deal with the cases that few or no failure time data could be obtained but degradation data of the primary parameter of the product are available. Step-Stress ADT (SSADT) is commonly used for the advantage that it needs only a few test samples to conduct a life test. For reliability and lifetime evaluation in SSADT, previous works use deterministic functions to represent the product performance degradation process. However, it does not represent performance degradation information adequately. It is necessary to add stochastic information description to performance degradation process. Time series analysis can represent stochastic information. During the last two decades, considerable research has been carried out in time series analysis. However, only few papers have studied the degradation data analyze method based on time series method. Moreover, SSADT data analysis based on time series method has not been reported in literature at present.

Optimization of the test stress levels of an ADT

In order to predict the reliability of the product with high reliability and long life, the accelerated degradation test (ADT) is commonly applied. However, in the studies of optimizing the ADT plans, there is nearly no researches on how to select the test stress levels. In this paper, the drift Brownian motion is selected as the degradation model. The op timization of the selection of the stress levels in a step stress accelerated degradation test (SSADT) is studied. The objective is to minimize the mean square error (MSE) of the prediction of the product operation reliability under the cost constraints. Through a Monte Carlo simulation method, the optimal stress levels and related sample size and test time are obtained. At last, the robustness of the results is shown through the sensitive analysis.

CSADT life prediction based on DAD using time series method

For long lifetime and high reliability products, it is difficult to obtain failure time data in a short time period. Hence, Accelerated Degradation Testing (ADT) is presented to deal with the cases that no failure time data could be obtained but degradation data of the primary parameter of the product are available. At present, there are mainly two ways to predict product life and reliability by ADT: one is based on degradation path, that is, product life prediction is obtained by prediction of each sample degradation path; the other is based on Degradation Amount Distribution (DAD), that is, product life prediction is obtained by prediction of all samples DAD parameters. Most previous works use deterministic model to represent the degradation path or parameters of DAD. However, long-term life prediction must take into account the stochastic and periodic nature of environmental variables. A few literatures study ADT life prediction using time series method for its excellent capable of stochastic and periodic information mining. However, life predictions using time series method in present literatures are all based on degradation path. Due to several special advantages of life prediction based on DAD, such as it can be used in random failure threshold situation, which is common situation in practice, it is important to study ADT life prediction based on DAD using time series method.

Life and reliability forecasting of the CSADT using Support Vector Machines

Accelerated Degradation Testing (ADT) is now adopted frequently to verify the reliability and life of high-reliable, long-life product. But ADT data analysis methods are still deficiency. Due to the excellent capable of little sample learning and nonlinear mapping, SVM prediction model is widely used in many fields. In this paper, a new degradation prediction method based on Support Vector Machines (SVM) is proposed and developed to predict time-to-failure of product. This prediction method is also compared with BPANN and regression methods to validate its effectiveness. Moreover, Constant Stress ADT is studied and ADT data are divided into several sets of performance degradation under different stress levels. Using SVM prediction method, all degradation processes are predicted to failure and lifetimes are obtained easily, then life and reliability under normal condition are evaluated by accelerated model. Simulation case demonstrates that the life and reliability prediction for CSADT based on SVM is reasonable and validity

Life-prediction of the CSADT based on BP algorithm of ANN

One of the main problems encountered when we study the reliability of high-quality, long-life products by traditional means is that it is difficult to obtain adequate failure data in the life testing, for example, cost and time are always limited. For many of these products, an underlying degradation process is the root cause of failures. In order to get useful information in a short time, accelerated degradation testing (ADT) is frequently used. Using multi-stress in the ADT, we not only reduce time and cost of the testing, and increase efficiency, but we also simulate the actual environmental conditions more accurately and obtain more credible results. However, to achieve these results, the accelerated model must be established. Unfortunately, it is often quite difficult to give a certain physical or chemical model and quantify the degradation process, because the failure mechanisms of different stresses are various. In this paper, a new model is developed to predict the life of items in the constant stress accelerated degradation testing (CSADT) based on Back-Propagation (BP) Algorithm of Artificial Neural Network (BPANN). With this BPANN model, unlike other degradation analysis methods, this acceleration model avoids complicated calculations. It provides a new approach to the life-prediction of the ADT.

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)

State tree analysis of FOG based on drift brownian motion

The binary state is the fundamental assumption to the traditional FTA. Hence, the traditional FTA gives only the binary logical analysis of product catastrophic failures. However, performance of Fiber Optical Gyroscope (FOG) will degrade with the time and it is able to perform its task with partial performance. To overcome this problem, multistate reliability analysis is presented. However, multistate reliability analysis is conducted by assuming that products have a number of discrete states. In most cases, product physical parameters are continuous real variables. Consequently, we combine the state analysis to FTA, and construct the state tree of FOG. By assuming that the state of SLD can be depicted by the drift Brownian Motion (DBM), we calculate the top event probability. Finally, we contrast the two top event probabilities of DBM and the model in [1], respectively.

Optimization of step stress accelerated degradation test plans

In the reliability prediction of the product with long life and high reliability, the step stress accelerated degradation test (SSADT) is commonly applied. With the motivation of predicting the product reliability most precisely, the problem of optimizing the test plans has drawn a lot of attentions in the application of SSADT. In this paper, the drift Brownian motion is selected as the degradation model. And in order to minimize the mean square error (MSE) of the prediction of the product operation reliability, the test plans of a SSADT that under the specified total test time and sample size are optimized through a Monte Carlo simulation method. At last, in combination with the sensitive analysis, the robust optimal test plans are obtained.