A distributed inference method integrating causal analysis and surrogate models for optimizing tuned mass damper parameters to enhance offshore wind turbine safety

Abstract

This study proposes a Distributed Inference Method integrating Causal Analysis and Surrogate Models (DIM-CASM), which aims to predict the structural responses and damage of offshore wind turbines (OWTs) under varying environmental conditions while adaptively optimizing Tuned Mass Dampers (TMDs) to enhance structural safety. As a novel paradigm for surrogate model application, DIM-CASM leverages its unique distributed inference model structure and deployment approach to fully utilize indirect variables often ignored in traditional surrogate models. The logic of data transmission within DIM-CASM closely mirrors the load transfer and interference processes in OWTs, resulting in higher fitting accuracy. In a case study involving a 15MW OWT, DIM-CASM was used to identify TMD parameters that adapt to varying wind and wave conditions. Compared to traditional fixed-parameter TMD strategies, the adaptive TMD derived through DIM-CASM demonstrates significant improvements in reducing the tower structure's extreme responses and coefficient of variation. Moreover, the DIM-CASM framework exhibits a unique interpretability advantage, enabling new mechanistic explanations through comparative analysis of pre- and post-optimization causal graphs. This reveals that the TMD's role might involve disrupting the influence of vertical structural responses at the tower top on the longitudinal responses, providing valuable insights for future studies on optimizing structural control strategies.