The invention discloses a modeling and on-line monitoring method based on multimodality collaboration time frame automatic division, which takes variation of process characteristics in batch shaft and time direction as well as time sequence of time frame running into consideration, and includes the steps of diving all the modalities through collaboration time frame, obtaining a unified time frame division result among different modalities, and building a uniform time frame model for similar process characteristics in each modality to simplify the modeling complexity. According to the invention, based on the result of collaboration time frame division, the relative change among all the modalities is analyzed, a multimodality statistic model based on different fluctuation types is built and applied in on-line monitoring, and the on-line monitoring performance is improved; the method is easy to implement, is successfully applied during the injection molding, not only facilitates the understanding of process characteristic, but also enhances the reliability and confidence level of actual on-line process monitoring, facilitates judging the running state of the industrial process, and timely discovers faults, thereby guaranteeing the safe and reliable running of actual production and the pursuit of high-quality products.