To address the challenge of accurately quantifying and predicting muscle fatigue, this study proposes a method for measuring muscle fatigue levels during dumbbell curl exercises. First, the Hill muscle model was used to calculate the biceps brachii power output during dumbbell curls for 12 male subjects. The true value of average muscle power output required under the current physiological state was fitted based on the average power output during non-fatigued cycles. The variation pattern between the curvature constant of the critical power (CP) model and the true muscle power output was investigated, leading to the construction of an improved CP model capable of adapting to physiological states. Next, residual muscle energy was calculated using a muscle energy expenditure and recovery model to quantify fatigue levels. Test results demonstrated that under varying exercise intensities, the mean root mean square error of the improved CP model in predicting the duration of dumbbell curls compared to actual measurement time is 8.01 seconds, significantly lower than the 19.79 seconds of the original model, validating the effectiveness of this approach. In summary, this methodology provides scientific basis for fitness enthusiasts to plan repetition counts within sets and rest intervals between sets during dumbbell curls. It plays a positive role in enhancing exercise efficiency and lays a foundation for further advancing the popularization of fitness activities.