In a variety of foundation treatment projects, the compaction degree of filler is one of the main quality control indicators. However, limited by the sparsity of measurement points and the requirements for implantable sensors, the existing test methods are difficult to achieve full-site visualization and rapid measurement. Therefore, this project takes the artificial vibration drum as the research object, takes the artificial neural network as the foundation, establishes a set of artificial neural network model which takes the artificial vibration as the foundation, and takes this as the foundation, establish a set of artificial neural network model based on artificial neural network. A new linear regression analysis method of drum vibration spectrum and amplitude characteristics based on multivariate characteristics is proposed, and a new packing compaction detection is realized. Through the measurement of the vibration frequency and amplitude of the rotary drum, and according to the physical properties and compaction characteristics of the soil, the accurate prediction is carried out. And its application prospect is discussed.