机械微弱信号自适应互补随机共振滤波增强检测与故障诊断研究

Condition monitoring and fault diagnosis are crucial in guarantee of machinery safety operation. The machinery incipient fault signal is rather weak, which brings difficulties in signal detection and fault diagnosis. Stochastic-resonance (SR)-based filter has been adopted in machinery signal enhancement and fault diagnosis due to its distinct advantage characterized by the fact that the weak periodic signal can be enhanced by proper noise. However, the flexibility and accuracy of SR-based filter are insufficient, which reduces effectiveness and efficiency in weak signal detection and fault diagnosis. Focusing on these issues, this project intends to investigate a new adaptive complementary SR-based filter which utilizes the sound and the vibration signals simultaneously to adaptively enhance the weak signal for fault diagnosis. The project includes four parts as: 1) construct an adaptive SR-based filter, and investigate the weighted summation approach of the feedback criterions based on the artificial neural network; 2) construct an objective function based on the synthetic feedback criterion, and investigate the genetic-algorithm–based multi-parameters optimization method; 3) investigate a dual-channel complementary SR-based filter in which the vibration and the sound signals are fused to enhance each other at the data-layer; and 4) study the diagnosis conclusion fusion method based on the Dempster-Shafer evidence theory at the decision-layer. This project expects to provide theoretical method for high-accuracy, high-efficient machinery condition monitoring and incipient fault diagnosis.

状态监测和故障诊断对于保障机器设备运行安全至关重要。机械早期故障信号十分微弱，对微弱信号检测是个难点，也是实现精密诊断的关键点。随机共振滤波器能利用噪声增强微弱周期信号而被应用于机械故障诊断中，但其适应性、实时性和精准度不足导致难以实现高效检测诊断。针对此问题，本项目拟研究一种新的自适应互补随机共振滤波器，通过声音和振动信号的互补融合滤波，实现机械微弱信号的自适应增强和故障诊断。具体内容包括：1）建立自适应随机共振滤波器模型，利用人工神经网络研究多反馈指标的加权组合方法；2）构建基于自适应反馈综合指标的目标函数，研究滤波器多参数遗传算法优化方法；3）建立双通道信号互补随机共振滤波器模型，研究利用振动和声音信号在数据层互补融合信号增强检测方法；4）利用DS证据理论研究双通道信号诊断结论的决策层融合方法。通过项目研究期望为实现高精度高效率的机械状态监测与早期故障诊断提供新的理论支持。

微弱信号检测是实现高效高精度机械故障诊断的关键点和难点。随机共振是一种能够利用噪声增强微弱信号的非线性系统。本项目深入研究了基于随机共振的微弱信号检测方法以及相关机械故障诊断方法，取得了一系列的创新成果并超额完成了预期的研究任务。代表性的创新成果简介如下：1）提出一种多指标融合的自适应随机共振方法，能够更全面地评估随机共振滤波器的性能并完成系统多参数的自适应寻优，从而实现机械微弱信号的自适应增强和诊断。2）提出一种双通道互补随机共振架构，将传统单通道信号输入拓展为双通道互补输入，双输入信号能够利用相关噪声实现互补增强，从而提高了微弱信号检测的效果。3）提出一种微弱声音和振动信号互补融合的随机共振架构，利用同源声音和振动信号的互补信息协同增强提高机械故障诊断的精度。研究成果共发表SCI期刊论文21篇、EI期刊论文1篇，其中项目负责人一作/通信论文19篇，包括振动声音信号处理和机械故障诊断领域权威期刊《Journal of Sound and Vibration》6篇、《Mechanical Systems and Signal Processing》2篇；机电系统和仪器测试领域权威期刊《IEEE Trans. on Power Electronics》、《IEEE Trans. on Instrumentation and Measurement》、《IEEE Sensors Journal》共8篇、《Measurement Science and Technology》4篇；《振动与冲击》1篇；申请/授权发明专利5项。上述文章在谷歌学术中被引用350余次，在WOS中SCI他引230余次，最高单篇引用78次。