设备在运行中萌生的故障(即早期故障)，特征信息微弱且往往被机械设备运行过程的强噪声所淹没，给故障诊断与预示带来困难，已成为国内外此领域研究的热点和难点。文章深入研究了机械故障动态信号与基函数的内积变换原理；提出了若干自适应多小波基函数构造方法；改进了几种多小波邻域区间和局部阈值降噪方法。利用典型的工程案例分析和阐述了重油催化裂化装置、连铸连轧机组、空分机、电力机车和船载卫星通信地球站传动系统在运行状态下，微弱动态信号的特征增强和复合故障特征提取的工程应用实效。

The faults initiated in operation (i.e. incipient fault) with the obscure symptoms and weak features, are always contaminated by a large amount of background noise. Hence, fault diagnosis and prognosis of incipient faults have been the difficulty and focus of the research field. The paper studied the principle of inner product transform of dynamic signals and basis functions, proposed several construction methods of adaptive multiwavelet basis functions,and improved several multiwavelet denoising methods with neighborhood and local threshold. The typical engineering cases of the equipment of heavy oil catalytic cracking, the continuous casting and rolling mills, the air compressor, the electric locomotive and the transmission device of satellite comunication on ship were studied, and the results showed the effectiveness of enhancement of weak dynamic signals and features extraction of compound faults.