<p>Collaborative creation, an essential component of collaborative learning, offers a powerful means for achieving deep knowledge building(KB). Yet, conventional methods of learning analysis struggle to effectively capture the dynamic and comprehensive nature of the KB process in collaborative settings. To address this, the present study implemented three rounds of collaborative creation activities, utilizing process mining, epistemic network analysis, and quantitative content analysis from a "Behavior-Cognition-Emotion" triadic perspective. This approach enabled an in-depth exploration of discussion patterns across different phases, among groups, and among individual learners throughout the collaborative creation process. The goal was to uncover learners' implicit behaviors, cognitive processes, and emotional dynamics, identify the evolutionary trajectories of their KB, analyze the internal influence mechanism of three dimensions of behavior, cognition and emotion, thereby uncovering key underlying patterns. The findings indicate that KB in collaborative creation progresses from independent shallow exploration to collaborative deepening. Common patterns emerge among different groups of learners, but high-performance groups will touch on the deeper KB. Within the same group, learners of different levels exhibit hierarchical differences in KB, and high-level learners usually act as leaders in the group. In addition, emotion plays a mediating regulatory role in the three dimensions of KB, in which confused emotion is an important bridge that drives KB from shallow to deep. Understanding these patterns of KB can aid educators in identifying potential issues within the collaborative process, enabling timely adjustments to teaching strategies to better support knowledge development.</p>