Exponential-polynomial divergence based inference for nondestructive one-shot devices under progressive stress model
Shanya Baghel, Shuvashree Mondal · 2025
Read Exponential-polynomial divergence based inference for nondestructive one-shot devices under progressive stress model by Shanya Baghel, Shuvashree Mondal (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-06414.txt.
Introduction
Shanya Baghel and Shuvashree Mondal built a robust estimation method for one-shot device testing in two thousand twenty-five. One-shot devices, like some aerospace and medical parts, can be tested only once, and they are checked under rising stress. Standard likelihood methods bend easily when data are contaminated.
Their approach uses a new exponential-polynomial divergence with three tuning settings, proven more flexible than a common rival. They study its large-sample behavior and its robustness through the influence function. A score matching method picks the tuning settings, balancing robustness against efficiency, and simulations compare it with older choices. They also design an optimal test plan under budget and time limits.
Catalog record
| CREATOR | Shanya Baghel, Shuvashree Mondal |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 69 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-06414
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Exponential-polynomial divergence based inference for nondestructive one-shot devices under progressive stress model Shanya Baghel, Shuvashree Mondal Abstract Nondestructive one-shot device (NOSD) testing plays a crucial role in engineering, particularly in the reliability assessment of high-stakes systems such as aerospace components, medical devices, and semiconductor technologies. Accurate reliability prognosis of NOSD testing data is essential for ensuring product durability, safety, and performance optimization. The conventional estimation methods like maximum likelihood estimation (MLE) are sensitive to data contamination, leading to biased results. Consequently, this study develops robust inferential analysis for NOSD testing data under a progressive stress model. The lifetime of NOSD is assumed to follow Log-logistic distribution. The estimation procedure addresses robustness by incorporating Exponential-polynomial divergence (EPD). Equipped with three tuning parameters, EPD based estimation is proven to be more flexible than density power divergence estimation frequently used for one-shot device testing data analysis. Further, we explore the asymptotic behaviour of minimum EPD estimator (MEPDE) for large sample size. The robustness of MEPDE is analytically studied through influence function. Since tradeoff between efficiency and robustness of EPD based estimation is governed by three tuning parameters, a novel approach leveraging Concrete Score Matching (CSM) is introduced to optimize the tuning parameters of MEPDE. Moreover, a comparative study with the existing methods of finding tuning parameters is conducted through extensive simulation experiment and data analysis. Another aspect of this study is determining an optimal plan to ensure a successful ALT experiment within specified budget and time constraints. It is designed on A-optimality criteria subject to the given constraints and is executed using the constraint particle swarm optimization (CPSO) algorithm. 1 Introduction Nondestructive one-shot device (NOSD) testing is a critical tool in engineering, particularly in the reliability prognosis of high-stakes systems such as aerospace components [ 1 ] , medical devices [ 2 ] , semiconductor technologies [ 3 ] and automotive safety systems [ 4 ] . NOSD can withstand multiple tests, offering additional data for reliability analysis, contrary to one-shot devices that are immediately destroyed after testing. Units such as brake or air conditioning fluids in cars, fan condensers, metal fatigue, light bulbs, circuit breakers, pressure relief valves and stability of pharmaceuticals are examples of NOSD [ 5 ] . For NOSD, often observations are limited to recording whether a failure occurs before or after a specified inspection time, resulting in the analysis of dichotomous data only. Furthermore, highly reliable products have become the standard due to the competitive market environment. To perform reliability analysis of such products within a specified time and budget, accelerated life testing (ALT) experiments are conducted; readers may refer to [ 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 ] and references therein. ALT can be classified according to the stress incorporation as constant stress, step-stress, progressive stress and cyclic stress, one may see [ 17 ] for more information on it. Zhuang et al. [ 18 ] stated that among these methods, progressive stress ALT (PSALT) is the most effective and flexible in reducing the life span of any product. In PSALT, the stress on test units increases continuously over time. For example, in bridge load testing, engineers progressively increase the weight or load on a bridge to determine its maximum capacity and identify potential failure points [ 19 , 20 , 21 ] . In aircraft wing stress testing, the wings of an aircraft are subject to progressive stress to simulate extreme conditions and ensure they can withstand real-world forces [ 22 , 23 ] . PSALT with linearly increasing stress is referred to as ramp-stress ALT. For more information on PSALT, readers may refer to [ 24 , 25 , 26 , 27 , 28 ] and references therein. Existing studies on the reliability analysis of NOSD are concentrated towards step-stress ALT (SSALT) experiments as evidenced by works of [ 29 , 30 , 5 , 31 ] and references cited therein. However, the reliability prognosis of NOSD under PSALT remains unexplored. The lifetime of NOSD testing units are commonly modeled by exponential, Weibull or lognormal distribution in the literature. There exist numerous real-world occurrences in which the lifetime of units follows a distribution different from these widely used ones. For instance, Log-logistic lifetime distribution is also used in survival analysis to model ALT data [ 32 ] . One can refer to [ 28 ] and references therein to go through the utility of Log-logistic distribution in modelling lifetime data. Its probabilistic behaviour resembles generalized Pareto distribution and Burr distribution [ 33 ] . In this study lifetime of NOSD is assumed to follow Log-logistic distribution since it is pretty versatile as its hazard rate can be decreasing or inverted bathtub-shaped [ 34 ] . When the failure mechanism of the data aligns with the assumed model assumptions, conventional maximum likelihood estimation (MLE) performs efficiently. However, its performance deteriorates significantly in the presence of contaminated data, where the failure mechanisms of some NOSD testing units slightly deviate from the assumed model. Since purely uncontaminated data is rare in practice, it is crucial to develop robust estimation methods that remain stable and accurate even when the data contains contamination. Such methods are essential for ensuring robust reliability prognosis in real-world scenarios, where deviations from model assumptions are inevitable. It is deciphered from the literature that density power divergence (DPD) [ 35 ] based estimation has been frequently exploited for robust inferential analysis; readers may refer to [ 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 ] and references cited therein. The tuning parameter \gamma in DPD based estimation (DPDE) coordinates robustness and efficiency. A higher value of \gamma enhances robustness by reducing sensitivity to outliers, while a lower value improves efficiency under standard model assumptions. Mukherjee et al. [ 44 ] introduced the Bregman-Exponential Divergence (BED) with the tuning parameter \alpha wherein the corresponding estimating equation can be interpreted as a weighted likelihood estimating equation. Furthermore, Singh et al. [ 45 ] introduced a generalized divergence known as Exponential-polynomial divergence (EPD) that encompasses both DPD and BED as special cases. The tuning parameters (\alpha,\beta,\gamma) in EPD enable greater flexibility in adjusting robustness and efficiency than DPD and BED. Therefore, this study proposes incorporating EPD for a robust estimation procedure for NOSD testing units. Additionally, the asymptotic distribution of minimum EPD estimate is thoroughly developed based on the discussion of Calvino et al. [ 46 ] . In conjunction with this, the influence function analysis is carried out to assess the robustness of the MEPDE analytically, providing deeper insights into its stability and performance. The tuning parameters (\alpha,\beta,\gamma) in EPD based estimation are crucial in balancing robustness and efficiency. These parameters vary across a continuous range, allowing countless alternatives in the estimation procedure. However, in any real-world scenario, the experimenter must select most suitable value of these tuning parameters for a specific dataset. Thus, finding optimal tuning parameters is very essential for such studies. In this direction, method proposed by Warwick and Jones (WJ) [ 47 ] and its improved version iterative Warwick and Jones (IWJ) algorithm [ 48 ] are frequently used. Castilla and