Intelligent self-adaption of marine engine degradation based on a digital-twin model

Abstract

Marine engine performance would be degraded due to component wear and tear during long-term operation. The degradation might be further serious if the marine engine is fueled with low/zero-carbon fuels. This research introduces an innovative adaptive digital twin (DT) framework to predict engine performance degradation with control shift of fuel injection and air intake system, due to nozzle hole wear, fuel supply system malfunctions, and valve train wear. The framework is confirmed by a model-in-the-loop system with three steps: benchmark engine modeling, degradation simulation and inverse solution by NSGA III. In the first step, a second-order response surface model (RSM) with a radial basis function (RBF) kernel is developed as a benchmark DT model, using various engine performance data. Since the high cost of marine engine experiments, the engine data was enlarged based on a sophisticated engine model. This model is then used in the second step to perform as a degrading engine by changing some of the parameters. This would cause performance deviations between the benchmark DT model and the degrading engine. In the third step, the NSGA-III algorithm is utilized to compensate the inputs of benchmark DT model, so that those performance deviations can be adaptively corrected. The results indicated that input offsets are identified with an error less than 4% and outputs deviations are predicted with relative error less than 0.9% between the corresponding DT model and degrading engine, which offers a reliable solution for accurate performance monitoring and control.