首先利用小波分析将非平稳随机信号进行分解，然后再由Ｆｏｕｒｉｅｒ变换对分解后的信号进行频谱分析，最终根据各频谱特征，实现对变速机械的故障诊断．以某冷轧厂新增３５００ｔ／ｈ大型冷却塔用变速及冷却系统的故障诊断为例，验证了本文提出的用小波分析对变速系统进行故障诊断的有效性 Firstly, the non-stationary random signal is decomposed by wavelet analysis, and then the spectrum of the decomposed signal is analyzed by Fourier transform. Finally, according to the characteristics of each spectrum, fault diagnosis of variable speed machine is realized. Taking the fault diagnosis of the variable speed and cooling system of 3500t / h large cooling tower in a cold rolling mill as an example, the validity of the wavelet analysis for fault diagnosis of the variable speed system is verified