一种兼具强干扰背景噪声特征的轴承故障数据模拟生成方法

Bearing fault data simulation generation method with characteristics of strong interference background noise

在对燃气涡轮发动机的轴承故障进行智能诊断时,存在着智能诊断难度大、实测案例稀缺等问题。为此,在考虑了非线性接触碰撞激励、发动机薄壁机匣支承结构复杂传递特性的基础上,提出了一种兼具发动机强干扰背景噪声特征的轴承故障数据模拟生成方法,并将其与实验数据进行了对比,验证了两者故障特征的一致性。首先,建立了典型发动机高压转子-轴承系统多体接触瞬态动力学模型,通过模拟故障轴承复杂接触碰撞过程,获取了非线性轴承故障激励力;其次,基于该发动机机匣高保真有限元模型,计算了故障轴承位置到测点的脉冲激励响应,并结合模态振型探明了发动机机匣的复杂传递特性;然后,在此基础上提出了基于非线性轴承故障激励与复杂传递的机匣测点信号时域离散卷积的重构方法,模拟了轴承故障在机匣测点的关键机理特征,并进一步融合了健康发动机的实测振动数据,构造了包含工作状态气流激振、非线性干扰等因素影响下的轴承故障仿真数据;最后,为了验证仿真数据故障特征的合理性,利用基于特征传递机理的包络解调方法提取了微弱轴承故障特征,并将其与发动机实测轴承故障实验数据进行了比较。研究结果表明:发动机轴承故障仿真数据与实验数据时域特征相似,二者包络谱中的故障特征频率误差小于1%,且故障特征频率及其倍频幅值大小相近,证明发动机轴承故障的仿真数据故障特征具有一定的合理性,可为基于机匣测点的发动机轴承故障特征提取、智能诊断算法训练等提供支撑。

In the intelligent diagnosis of gas turbine engine bearing faults,there are some problems such as difficulty in intelligent diagnosis and scarcity of measured cases.Therefore,considering the nonlinear contact-impact excitation and the complex transmission characteristics of the engine thin-walled casing support structure,a bearing fault data simulation generation method with the characteristics of strong interference background noise of the engine was proposed.Comparing with the experimental data,the consistency of their fault characteristics was verified.Firstly,a multi-body contact transient dynamics model of a typical engine high-pressure rotor-bearing system was established,and the nonlinear bearing fault excitation force was obtained by simulating the complex contact and collision process of the fault bearing.Secondly,based on the high-fidelity finite element model of the engine casing,the pulse excitation response from the fault bearing position to the measuring point was calculated,and the complex transmission characteristics of the engine casing were explored in combination with the modal vibration shape.Then,on this basis,a time domain discrete convolution reconstruction method of casing measuring point signals based on nonlinear bearing fault excitation and complex transmission was proposed to simulate the key mechanism characteristics of bearing faults at casing measuring points.By further integrating the measured vibration data of healthy engines,the simulation data of bearing faults under the influence of factors such as working state airflow excitation and nonlinear interference were constructed.Finally,in order to verify the rationality of the fault characteristics of the simulation data,the envelope demodulation method based on the feature transfer mechanism was used to extract the weak bearing fault characteristics which were compared with the engine measured bearing fault experimental data.The research results show that the time domain characteristics of the engine bearing fault simulation data and the experimental data are similar,the fault characteristic frequency error in the envelope spectrum of the two is less than 1%,and the fault characteristic frequency and its multiplier amplitude are similar in size.It proves that the fault characteristics of engine bearing fault simulation data have certain rationality.It can provide support for engine bearing fault feature extraction and intelligent diagnosis algorithm training based on casing measurement points.