The friction factor between the drill string and the wellbore is critical for force analysis of drilling strings and drag-torque prediction. Currently, the friction factor between the drill string and the wellbore is primarily determined through empirical summaries or manual calibration of benchmark loads for inversion. However, the timeliness, accuracy and intelligence of these methods still need improvement. To address this, a closed-loop dynamic inversion method of friction factor between drill string and wellbore driven by data and mechanism is proposed, which mainly includes the following steps: ①Drilling window data are obtained based on the characterization patterns of drilling data under different drilling conditions; ②The density peaks clustering(DPC) algorithm is used to cluster data from the drilling window, whereby key parameters for the inversion of the friction factor(off-bottom torque and off-bottom hook load) are automatically extracted; ③Based on the boundaries at both ends of the drill string, the friction factor is inverted using the drag torque model; ④The drilling conditions are monitored in real-time, and steps ①~③ are repeated to achieve closed-loop dynamic inversion of friction factor. Taking well A1 as an example, the dynamic inversion of the friction factor was conducted. The calculation results show that the average error between the clustering results of DPC algorithm and the expert experience is 4.31 %, which meets the requirements of field operation. The proposed friction coefficient inversion method is accurate and reliable. The risk of drill string sticking can be predicted by the dynamic inversion results of the friction factor, which can also be used to correct the weight on bit and torque on bit in real-time to establish a more accurate artificial intelligence prediction model. The closed-loop dynamic inversion of the friction factor between the drill string and wellbore is realized by method proposed. The timeliness, accuracy and intelligence of friction factor calculation are enhanced, thereby providing solid theoretical support for intelligent drilling.