Nowadays online quality estimation for the resistance spot welding (RSW) has benefited a lot from monitoring the electrode displacement caused by nugget thermal expansion. Based on these emerging monitoring techniques a new approach is proposed to classify the weld quality and assure the quality for mass-produced weld group, which enables the continuous quality improvement concept during the welding process. A causal models are built with the offline trained Bayesian Belief Networks (BBN). It is a weld quality assessment net reveals the dependency of the weld quality on the features displayed by the displacement curve, which can be used for overdesigning the safety welds or as the probabilistic forecasting model for online weld quality assessment. The experimental results show that the proposed approach is valid and feasible to predict the weld quality and assure the overall quality for weld group in real applications.