Non-contact and full-field online monitoring of curing temperature during the in-situ heating process based on deep learning

Online monitoring of the curing temperature field is essential to improving the quality and efficiency of the manufacturing process of composite parts. Traditional embedded sensor-based technologies have difficulty monitoring the full temperature field or have to introduce heterogeneous items that could have an undesired impact on the part. In this paper, a non-contact, full-field monitoring method based on deep learning that predicts the internal temperature field of composite parts in real time using surface temperature measurements of auxiliary materials is proposed. Using the proposed method, an average temperature monitoring accuracy of 97% is achieved in various heating patterns. In addition, this method also demonstrates satisfying feasibility when a stronger thermal barrier covers the part. This method was experimentally validated during the self-resistance electric heating process, in which the monitoring accuracy reached 93.1%. This method can potentially be applied to automated manufacturing and process control in the composites industry.