With the advent of Industry 4.0, intelligent manufacturing and the industrial Internet are changing traditional production methods. In this paper, a multimodal industrial process monitoring method is proposed. By solving the accuracy and robustness of industrial process monitoring, two models are used to solve the problems of multimodal text alignment and adding noise for training to improve the stability, similarity and consistency of the system. At the same time, different types of data sources are integrated to achieve real-time, comprehensive and accurate monitoring of the production process. This method uses deep learning technology and combines multiple modes such as image recognition and sensor data analysis to build an efficient monitoring system.