A Novel Method to Improve Temperature Uniformity in Polymer Composites Microwave Curing Process through Deep Learning with Historical Data

Compared with processing methods using conductive heating, microwave processing technology has many advantages such as its extremely short processing time and low energy consumption. However, the uneven temperature on the composite surface resulted from the uneven electromagnetic field distribution have become a big problem. Because the traditional model-based approach was difficult to establish the relationship between the composite temperature behaviors and microwave control strategies, existing methods mainly alleviated this problem by generating a relative movement between the microwave field and the object being heated, which cannot essentially achieve a uniform temperature distribution due to the uncertainty of the random compensation principle. In this paper, a data-driven method was proposed to solve this problem using an optimized convolutional neural network with extensive historical data. On this basis, the monitored uneven temperature distribution on the composite surface was accurately compensated in real time. Experimental results indicated that a reduction of ~53% in temperature difference was achieved compared with existing methods.