Simulation-Based Transfer Learning for Support Stiffness Identification

Abstract

The mounting of a rotating machine affects the dynamic behavior of the machine. Typically in large machines, the support structures have lower stiffness on the actual site than in the acceptance tests conducted by the manufacturers. In this research, a method is developed for the support stiffness identification for an in-situ machine using a simulation-data-driven, deep learning algorithm. The novel approach aims to utilize transfer learning to first teach the deep learning algorithm using vibration response data generated from a simulation model of the rotor-bearing-support system, and then test it with measured response. To validate the stiffness estimation of the algorithm for multiple cases, an experimental test rig is used where the horizontal support stiffness can be varied through a range of values. The results from the deep learning algorithm are compared with simpler algorithms such as Linear regression (LR), Artificial Neural Network (ANN), and Support vector regression (SVR) for benchmarking. The models are trained with filtered frequency domain response, and challenges due to measurement uncertainty are analyzed. With proposed pre-processing steps of the frequency domain response and outlier elimination, the deep learningbased virtual sensor can predict the support stiffness with reasonable accuracy, where the limiting factor is the data quality and lack of excitation at critical speed frequencies.

Physics-Based Digital Twins Merging With Machines: Cases of Mobile Log Crane and Rotating Machine

“…The performance and durability of industry-scale structures can be evaluated by applying physics-based model-driven methods or data-driven measurement-based methods. Recent research has demonstrated that combining these approaches can result in excellent simulation performance and accuracy [17]. In principle, the accuracy of calculation models correlates with the level of detail built into the structural model, i.e., the number of elements used (fineness of the mesh).…”

“…Physics-based simulation enables developing virtual prototypes that are subject to real-life physical constraints [16] while real-time capability extends the use of simulation to further product lifecycle stages. Computationally effective dynamics modeling opens the possibility for, e.g., fault and state identification, problem root source debugging, and predictive maintenance [17]. State-of-the-art simulation models can also account for system hydraulics in real time [18] and therefore respond to user inputs as well.…”

Physics-based dynamic simulation opportunities with digital twins

“…The physics-based simulation models can produce data for that purpose. Especially including non-idealities and faults in the datasets can be created with ease [5].…”

Research on parameter identification method for the power transmission auxiliary system of tracked vehicles based on MOGA-LSTM

“…Yang has proposed a method for identifying the connection parameters using SVR and NSGA-Ⅱ algorithm [4] . Zhao et al has used the error function of the test and simulation results to identify the contact stiffness of the bolt connection structure of the whole circle drum [5] , Gu et al has presented a simultaneous identification method of all the stiffness components of orthotropic composite materials [6] .Bobylev et al has developed a method for the support stiffness identification for an in-situ machine using a simulation-data-driven, deep learning algorithm [7] . This paper takes one of the typical components of the power transmission auxiliary system of a tracked vehiclethe controller as the object, comprehensively considers accuracy and efficiency, and establishes a parameter identification model based on MOGA algorithm and LSTM network.…”