Order-Restricted Inference for Exponentiated Rayleigh Distribution Under Multiple Step-Stress Accelerated Life Test

Area of Science:

• Statistics
• Reliability Engineering
• Accelerated Life Testing

Background:

• Accelerated life testing (ALT) is crucial for estimating product lifetimes under normal use conditions by subjecting units to higher stress levels.
• Step-stress testing is a type of ALT where stress levels are progressively increased, accelerating failure and reducing expected lifetime.
• The exponentiated Rayleigh distribution and cumulative exposure model are often used to model lifetimes in ALT, but incorporating order restrictions is complex.

Purpose of the Study:

• To develop and compare both frequentist and Bayesian statistical inference methods for multiple step-stress accelerated life tests.
• To incorporate the inherent order restriction (increasing stress leads to decreasing lifetime) into the statistical analysis.
• To model unit lifetimes using a two-parameter exponentiated Rayleigh distribution related by the cumulative exposure model.

Main Methods:

• Frequentist approach: Employing reparameterization to derive order-restricted maximum likelihood estimates (MLEs) and constructing confidence intervals using the Fisher information matrix.
• Bayesian approach: Conducting Bayesian analyses and generating credible intervals through importance sampling techniques.
• Utilizing the cumulative exposure model to link lifetime distributions across different stress levels.

Main Results:

• The study provides methodologies for obtaining order-restricted estimates and intervals in both frequentist and Bayesian frameworks for step-stress ALT.
• Demonstrated the application of these methods through extensive simulation studies.
• Validated the proposed methods by analyzing a real-world dataset.

Conclusions:

• Both frequentist and Bayesian methods effectively handle order restrictions in multiple step-stress ALT under the specified distributional assumptions.
• The developed techniques offer robust tools for reliability analysis in accelerated testing scenarios.
• The findings contribute to more accurate lifetime predictions and reliability assessments for products subjected to varying stress conditions.