Development of a wind turbine gearbox virtual load sensor using multibody simulation and artificial neural networks

Abstract

A current development trend in wind energy is characterized by the installation of wind turbines (WT) with increasing rated power output. Higher towers and larger rotor diameters increase rated power leading to an intensification of the load situation on the drive train and the main gearbox. However, current main gearbox condition monitoring systems (CMS) do not record the 6‑degree of freedom (6-DOF) input loads to the transmission as it is too expensive. Therefore, this investigation aims to present an approach to develop and validate a low-cost virtual sensor for measuring the input loads of a WT main gearbox. A prototype of the virtual sensor system was developed in a virtual environment using a multi-body simulation (MBS) model of a WT drivetrain and artificial neural network (ANN) models. Simulated wind fields according to IEC 61400‑1 covering a variety of wind speeds were generated and applied to a MBS model of a Vestas V52 wind turbine. The turbine contains a high-speed drivetrain with 4‑points bearing suspension, a common drivetrain configuration. The simulation was used to generate time-series data of the target and input parameters for the virtual sensor algorithm, an ANN model. After the ANN was trained using the time-series data collected from the MBS, the developed virtual sensor algorithm was tested by comparing the estimated 6‑DOF transmission input loads from the ANN to the simulated 6‑DOF transmission input loads from the MBS. The results show high potential for virtual sensing 6‑DOF wind turbine transmission input loads using the presented method.

Sensor Screening Methodology for Virtually Sensing Transmission Input Loads of a Wind Turbine Using Machine Learning Techniques and Drivetrain Simulations

“…On the other hand, solutions that aim to indirectly or virtually sense transmission input loads face a different challenge of collecting sufficiently representative data during WT operation to develop and validate their solutions [ 8 ]. Azzam et al addressed this need by providing a methodology for generating the necessary data for virtual load sensor development using multibody simulations (MBS) of a Vestas V52 WT model subjected to simulated wind fields according to IEC 61400-1 [ 9 ]. The authors demonstrated that MBS simulations can be used to significantly reduce the resources needed during development of WT virtual load sensors by providing a more flexible, less costly alternative to physical measurement campaigns as a data source for virtual sensor prototyping.…”

“…In this investigation, we address this need by presenting a methodology for sensor screening during the data-scarce virtual sensor design phase prior to real-world testing. Using the methodology recently published by Azzam et al [ 9 ], simulated wind fields covering a variety of operating conditions were applied to an MBS WT model to generate a time series of displacements, misalignments, and other parameters at various locations in the WT drivetrain. Using these simulated data, the methodology presented in this paper is intended to prioritize candidate sensors prior to real-world testing in order to (1) reduce the number of sensors required and (2) enhance the utility of the subsequent instrumentation and real-world testing phases, which were not repeatable in this project due to resource constraints.…”

“…In order to generate the time-series dataset of the 6-DOF GB input loads (hereafter referred to as “target variables”) and the resulting deformations, misalignments, and rotational speeds of drivetrain components (hereafter referred to as “predictor variables”) to be analyzed in this investigation, Azzam et al, assembled models of the various WT components in a virtual MBS environment [ 9 ]. The rotor blades, as well as the tower, were modally decomposed after an initial modelling in finite elements (FE) [ 21 , 22 ].…”

“…The rotor blades, as well as the tower, were modally decomposed after an initial modelling in finite elements (FE) [ 21 , 22 ]. In addition, various drivetrain components, e.g., main frame and gearbox planet carrier, which contribute to the predictor variables, were modeled as flexible bodies in the MBS environment, whereas micro-level flaws, such as material imperfections, were not incorporated in the simulations [ 9 ]. In order to subject the WT model to simulated wind fields, TurbSim, together with the AERODYN force element and a SIMULINK PI controller, was used as part of a co-simulation with MATLAB [ 23 , 24 ].…”

“…In order to subject the WT model to simulated wind fields, TurbSim, together with the AERODYN force element and a SIMULINK PI controller, was used as part of a co-simulation with MATLAB [ 23 , 24 ]. In their paper on the topic, Azzam et al provide a more in-depth explanation of the simulation model, as well as a discussion of its validity with respect to the task of virtually sensing WT transmission input loads [ 9 ].…”