深海载人潜水器推进器系统故障诊断的新型主元分析算法
Principal component analysis algorithm for fault diagnosis of thruster system in dep-sea human occupied vehicle

针对“蛟龙号”多推进器系统故障检测与快速定位难题，将信度分配模糊小脑神经网络FCA-CMAC(Credit Assigned Fuzzy Cerebellar Model Articulation Controller)应用于主元分析模型，提出一种基于主元分析(Principal Component Analysis,PCA)的深海载人潜水器推进器系统故障诊断模型。首先，应用推进器系统正常运行的历史电流样本数据，由主元分析模型得到各推进器的电流预测值。其次，计算出故障检测统计量均方预测误差（Squared Prediction Error,SPE），根据SPE值是否跳变，判断推进器系统有无故障发生。通过分别重构各推进器电流信号的SPE值对故障推进器进行定位和隔离。最后，通过对实际海试数据进行仿真处理说明了该算法的可行性，并通过与BP(Back Propagation)神经网络和常规CMAC(Cerebellar Model Articulation Controller)神经网络进行比较，说明基于FCA-CMAC的主元分析模型的优越性。
For the problem of fault detection and fault isolation in the multi-thruster system, a fault diagnosis model of thruster system in deep-sea human occupied vehicle based on principal component analysis(PCA) and fuzzy cerebellar model articulation controller neural network (FCA-CMAC) is proposed.Firstly,the forecasting electric current values of thrusters are computed by using historical data measured under fault-free conditions and the PCA model.Secondly, the squared prediction error (SPE) is calculated to characterize the operational status of the thruster system. A fault can be detected when the SPE increases suddenly.Current values are reconstructed respectively to newly calculate the SPE to locate the faulty thruster. Finally, compared to BP and conventional CMAC,the method proposed is proved feasible and effective by the simulation of the actual sea trial data.