基于VR与PNN相结合的机电液控制系统故障诊断方法

针对大型复杂机电液控制系统故障诊断中存在的数学模型获取困难、历史故障数据匮乏问题，提出了一种将虚拟样机与概率神经网络相结合的故障诊断混合方法。建立系统的虚拟样机，并对其可信性进行校核与验证。在此基础上进行大量随机性故障植入与仿真实验，获取故障仿真数据。经过特征提取与概率神经网络模式识别训练，形成用于诊断的知识库，从而实现故障诊断。以操舵系统作为研究案例，得到了较高的故障检测和隔离精度与较低的虚警及漏警率，验证了该方法的可行性，为大型复杂机电液控制系统故障诊断提供新的思路。
In the diagnosis of large-scale mechanic-electronic-hydraulic control system, and for the mathematical model was hard to build and the historic fault data was short, a hybrid fault diagnosis method based on virtual prototyping and PNN (probabilistic neural network) was proposed. Virtual prototyping was first built and its credibility was validated. On this basis, fault injection and simulation were conducted to obtain fault data, which was then extracted as fault features and trained by PNN to form diagnosis knowledge library. A case study of steering system was presented to verify the correctness of the proposed method, which shows that the accuracy of fault detection and isolation is high and the rate of false/missing alarm is low. The proposed method may bring a novel idea for the fault diagnosis of large-scale and complicated mechanic-electronic hydraulic control system.