Signal de-noising is one of the most important project in fault diagnosis and signal processing for rotating machinery. It could be carried out in a lot of ways. Wavelet de-noising, which is becoming popular recently, is a very promising method for this purpose. This article mainly introduced basis of wavelet theory and wavelet de-noising and its application. There are a lot of factors which will affect the de-noising efficacy, such as wavelet type, threshold strategy, decomposition level and so on. Some valuable rules was discovered by the analysis of simulated typical fault signals of rotating machinery.