Abstract
The ship structural stress monitoring system is one of the main technical approaches to realize intelligent ship hull structure, and it has been applied to high-performance ships and large-scale merchant cargo ships in recent years. The structure monitoring system can not only serve as the input of data-driven digital twin models of ships structures, but is also the basis for the derived decision support system. For this reason, it is crucially important to obtain accurate stress in the structure, and the error characteristics of strain gauges under possible combinations of load and temperature that a ship may undergo must be investigated. For a ship structure, the error characteristics of the strain gauges can be investigated by comparing the measures from the strain gauges installed on the ship and the stress obtained from numerical simulation. A neural network to compensate the errors of the strain gauges can be trained through measures from strain gauges and the numerical results for the stress at the same location. In this study, the analysis of the performance and the error characteristics of the strain gauges on a beam test piece are conducted. An experimental investigation of the response of two types of fiber Bragg grating strain gauge at different temperatures and different loads are conducted in the same approach. The performance and error characteristics of the strain gauges under different loads and temperatures are analyzed. Based on the analysis of error characteristics, various BP neural networks are constructed to compensate the errors of the strain gauges. Comparison of compensation results and experimental results of two types of fiber Bragg grating strain gauge shows that the proposed method can effectively reduce the influence of the error on the accuracy of the gauge. Application of this method requires the training samples of measures and numerical results for the stresses under known static loading conditions like berthed in a harbor or moving in calm water. Thus it is feasible to update the neural networks for the compensation of errors during the operation of ships.
