FEM modeling of hamburger pan cooking: Fat content influence and neural network-based prediction of fat loss

In this study, we developed an improved contact-cooking model that incorporates variations in fat content and its retention capacity, aiming to accurately simulate products with different compositions. The proposed approach incorporates the structural heterogeneity of meat by distinguishing between muscle fibres and interstitial fluid, and simulates the transport of water and fat between these regions. The model also accounts for heat transfer and meat deformation, representing the tissue as a hyperelastic material. The computational framework was implemented for the pan-cooking of hamburgers with different fat contents (ranging from 3% to 24%) and periodic flipping during the process. To validate the model, cooking experiments were performed. Although increasing fat content did not significantly affect water loss or core temperature ( p > 0 . 05 ), it strongly influenced other critical aspects of cooking performance and product quality, such as surface temperature, fat loss, total cooking losses, and shrinkage. These changes are relevant because they impact texture and consumer perception.In addition, hamburgers with a higher fat content exhibited lower hardness, cohesiveness, gumminess, and chewiness. The model successfully predicted these trends, demonstrating its potential to capture fat-related effects beyond thermal behavior and enabling the use of model-generated data for the training, validation, and testing of a simple neural network to predict fat loss during the cooking of hamburgers with varying water, fat, and protein contents. • Tests show higher fat content increases fat loss/shrinkage but barely affects water loss. • Fat-rich samples heat up more near pan, enhancing surface dehydration. • Validated cooking model accounts for meat composition, shrinkage and dynamics. • A trained neural net predicts fat loss from raw meat composition using a model.