<p>To address the issue in automated drilling and riveting equipment (ADRE) where unclear machining quality during batch production leads to the inability to accurately perceive and dynamically control machining errors, a dynamic error control system based on an error mechanism model is developed. This system consists of an advanced data flow framework and a dynamic error control method embedded within it. Leveraging a digital-thread-based cloud–edge collaborative architecture, the method predicts the riveting quality in real time based on machining parameters and dynamically compensates and optimizes the parameters for the subsequent rivet, thereby enabling precise and controllable error management. Within this method, an original error mechanism model for the ADRE system is proposed based on the system’s structural characteristics. Guided by this model and the minimum-parameter principle, a set of optimized machining parameters is selected for algorithm training. Furthermore, a riveting quality prediction algorithm with strong feature-capturing capability is developed. This algorithm named AttenBiNet integrates MSCNN modules and self-attention mechanisms for deep feature extraction, followed by a BiLSTM unit to perform quality prediction. The model exhibits strong robustness even in the presence of missing data. Finally, a dataset comprising real production data from 4712 rivets was constructed using self-developed ADRE equipment and utilized for training the proposed AttenBiNet. The results demonstrate that the optimized algorithm reduced the MAE by 66.83% and RMSE by 64.18%. Under data-missing conditions, the MAE increased by only 0.0045 and the RMSE by only 0.0093, confirming the effectiveness and robustness of the proposed dynamic error control method.</p>