Deep neural network-based multiagent synergism method of probabilistic HCF evaluation for aircraft compressor rotor

Abstract

The high cycle fatigue (HCF) of compressor rotor seriously affects the performance and reliability of aircraft engine. Probabilistic HCF evaluation is an effective measure to quantify the uncertain traits of vibration stresses and assess the reliable HCF life for compressor rotor. To improve the computing accuracy and efficiency of probabilistic HCF evaluation, a deep neural network with multiagent synergism (DNN-MS) is proposed by absorbing the strengths of the developed DNN model with flexible topology structure into the MS approach. The probabilistic HCF evaluation of a typical compressor rotor is considered as one case to evaluate the presented method. Analysis results reveal that the reliable HCF life of compressor rotor under reliability 99.87 % is 118,199 cycles, and the physical parameters (rotational speed, outlet static pressure, material density and elastic modulus) play a leading role on HCF life since their effect probabilities of 28 %, 19 %, 18 % and 17%, respectively. The comparison of methods (direct Monte Carlo simulation, quadratic polynomial, support vector regression, neural network and the presented DNN (unoptimized and optimized)) shows that: (1) the proposed DNN-MS possesses the best modeling accuracy among the agent model methods, with the mean absolute error (MAE) of 0.0176 and the root mean squared error (RMSE) of 0.0213; (2) the computing accuracy of DNN-MS is the highest among the agent model methods and almost consistent with direct Monte Carlo simulation; (3) DNN-MS holds the highest efficiency in probabilistic HCF evaluation.