Research on the unbiased failure rate model based on a nonlinear Wiener process and transfer learning

To resolve unbiased failure rate prediction challenges for aerospace components under small-sample scenarios—where conventional methods fail to capture nonlinear degradation patterns and parameter estimation bias. An unbiased failure rate prediction framework is proposed by integrating nonlinear Wiener process modeling, transfer learning, and failure mechanism consistency verification. The methodology establishes a nonlinear implicit Wiener process with state-space formulation to characterize degradation trajectories; implements variance-corrected Expectation-Maximization (EM) algorithm for unbiased parameter estimation; employs Semi-Supervised Transfer Component Analysis (SSTCA) and Support Vector Machine (SVM) for cross-domain degradation prediction; constructs variation coefficient-based hypothesis testing to validate physical mechanism consistency. The proposed framework's effectiveness and feasibility are validated through a failure rate evaluation case study on aircraft flap hinge joint bearings. Comparative analyses demonstrate the proposed method's superiority in Remaining Useful Life (RUL) and failure rate prediction accuracy over conventional approaches for rotating aerospace components. This framework offers a physics-informed solution for reliability engineering, effectively addressing data scarcity challenges in prognostic health management of safety-critical aerospace components.