Planer linkage mechanisms are always designed to realize a desired functional output or path. Due to the uncertainty from dimensional parameters and joint clearance, the mechanism motion would deviate from the desired motion. Effectively estimating the operational reliability of mechanisms is an important challenging issue in engineering applications. The current reliability methods only focus on the worst time point or several discrete time points, which may lead to large error due to ignorance of the time correlation. This paper proposes a timedependent reliability approach to predict the probability over a continuous time interval for planer linkage mechanism considering the uncertainty existing in mechanism dimensions and the joint clearances. Firstly, the motion error model considering the uncertainty in variables and joint clearances is introduced, in which the variables of mechanism dimensions are assumed to be independent and the variables of joint clearance are considered as dependent. Then the surrogate modeming method is employed to simulate the actual stochastic process and predict the time-dependent reliability. Training points of random variables, joint clearances and over time are generated. The surrogate model is refined adaptively based on the learning function and a developed convergence criterion. Results of a four-bar function generation mechanism are utilized to verify the validity and application of the proposed approach to demonstrate the accuracy and efficiency.