Bolt loosening is a common failure mode in bolted joints, and the quantitative assessment of bolt preload is crucial for structural safety maintenance and accident prevention. To address the limitations of conventional ultrasonic testing, which is susceptible to temperature interference and difficult to interpret, this paper proposes a bolt preload monitoring method based on coda wave interferometry (CWI). This method transforms ambient temperature from a conventional source of interference noise into a controllable active excitation source, actively leveraging the nonlinear effects induced by temperature. By comprehensively considering the coupled effects of thermal expansion and thermal softening, a nonlinear theoretical model is established between temperature variations and coda wave characteristic parameters (relative velocity change Δv/v and decorrelation coefficient Kd). Experiments are conducted on a copper-aluminum heterogeneous material bolted joint structure. A single-side transducer arrangement is selected to enhance signal stability. Temperaturecontrolled experiments and wave signal acquisition and analysis are conducted within the range of 22.5℃ to 31.0℃ in 0.5℃ increments for three torque states (15, 10 and 6 N·m), corresponding to the slow, fast, and severe loosening stages, respectively. The results indicate that the sensitivity of the coda waves to temperature variations is significantly higher than that of the direct waves. At 30℃, the Pearson correlation coefficients (PCC) of the coda waves under the three torque conditions decrease to 0.72, 0.41, and 0.05, respectively, with a maximum reduction of 93.06%, whereas the reduction for the direct waves is only 7.37%. The Δv/v exhibits a strong quadratic negative correlation with ΔT (R2>0.99), with its linear coefficient increasing as torque increases. The Kd exhibits a strong quadratic positive correlation with ΔT (R2>0.99), with its quadratic coefficient increasing as torque decreases, making it an effective characteristic parameter for distinguishing different degrees of bolt loosening. This study confirms the feasibility of using coda wave interference method to monitor the thermal softening effect of bolted joints, providing a new strategy based on active temperature excitation for bolt loosening monitoring in variabletemperature environments, thereby overcoming the limitations of conventional compensation methods.