The complex and coupling between electrical, thermal, and mechanical behaviors in cylindrical LiFePO4 batteries, create fundamental barriers to precise SOC estimation. Hence, in this study, constant current constant voltage-charge-constant current discharge experiments and dynamic operating conditions experiments are conducted under a wide temperature range for cylindrical LiFePO4 batteries at different aging stages firstly. Then, a fusion SOC estimation methodology is proposed by integrating electro-thermal–mechanical behaviors. Finally, SOC estimation are performed under simulating real-world onboard application scenarios. Experiments results reveal that both temperature evolution and strain evolution exhibit distinct patterns, with strain evolution demonstrating strong SOC dependence and charge–discharge asymmetry. Meanwhile, the aged battery exhibit more pronounced temperature fluctuations and strain fluctuations than the fresh one, and the mechanical behavior of the aged battery is being more susceptible to ambient temperature variations. SOC estimation results demonstrate that introducing the mechanical behavior is beneficial for significantly improving the SOC estimation accuracy, the mean absolutely error remains below 1.10% and the root mean square error remains below 2.00%. Compared with conventional data-driven approaches, the proposed methodology achieves an average reduction of 30.10% and 31.51% respectively.