The reliability of soft robotic devices will be the bottleneck that slows their commercialization. In particular, fatigue failure issues are a major concern. Thus, reliability should be taken into account from the earliest stages of development. However, to date, there have been no attempts to analyze the reliability of soft robotic devices in a systematic manner. When soft robots are employed to force transmission applications, reliability is typically a dominant issue, since soft robotic structures are constructed with soft material components; these materials have highly nonlinear properties that arise due to the large distribution in the material properties. Furthermore, reliability should be analyzed from the robot’s system down to the components using domain knowledge about the system; this requires a systematic approach. This study presents a framework for reliability analysis of soft robotic devices taking into account of a probability distribution that has not been considered before and examines a case study of a tendon-driven soft robot. This study focuses specifically on 1) concept design process, 2) lifetime analysis process, and 3) design and optimization process. A life model that considers distribution is proposed using accelerated life testing (ALT) based on analysis of the failure mechanism of the tendon-driven system. The tensile stress of the wire was varied during the experiment with different bend angles and output tension. The result was validated with different stress levels using a testbed to simulate an actual application. The proposed reliability analysis methodology could be applied to other soft robotic systems like pneumatic actuators to improve the reliability-related properties during the robot design stage, and the life model can be used to estimate the device lifetime under various stress conditions.