<p>Ultrasonic vibration-assisted cutting (UVAC) technology is widely applied in the processing of difficult-to-machine materials. To further improve processing quality and efficiency, adaptive control based on high-frequency dynamic cutting force has become a research focus. However, challenges are posed due to its high-frequency vibration disturbance and cutting force components in instantaneous dynamic monitoring. To address this issue, a dynamic monitoring system with high resolution is designed in this study based on polyvinylidene fluoride (PVDF) piezoelectric strain sensors to measure high-frequency three-dimensional UVAC forces. This system simplifies the strain rose structure of the sensor layout on the device. Based on this structure, the “force-to-charge” conversion equation is established by combining the constitutive equation of the PVDF sensor with Timoshenko beam theory, describing the relationship between the sensor’s output charge and the cutting force applied during the cutting process in ultrasonic devices. Meanwhile, A threshold function based on multi-scale variable factors and a threshold selection method based on Bayesian estimation are proposed. By incorporating prior information of noise and high-frequency vibration excitation characteristics, the method enables the effective separation of cutting force and high-frequency vibration signals. Finally, orthogonal cutting experiments were conducted for verification. The results indicate that the average UVAC forces measured by the proposed system in the <i>X</i>, <i>Y</i> and <i>Z</i> directions are in excellent agreement with those obtained from the dynamometer. For 304 stainless steel, the average errors are 6.40%, 7.56% and 6.69%, respectively, while for the titanium alloy and 2A12 aluminum alloy they are 5.90%, 5.64% and 6.64%, respectively. Furthermore, the system demonstrates better performance in monitoring high-frequency instantaneous cutting forces compared to the dynamometer.</p>