Physical-Information-Driven-and-Guided-Multimodal-Multi-task-Neural-Network

Physical Information-Driven and Guided Multimodal Multi-task Neural Network (PMNN)

Code Explanation:

'mutilcontrolValueAndStabilityModelUnetPINNRE.py'represents the model code of the PMNN.

'mutilcontrolValueAndStabilityModelUnetRENonPLoss.py'represents the model code obtained after PMNN removed the physical equations from the loss function.

'mutilcontrolValueAndStabilityModelUnetRENonInPhysic.py'represents the model code obtained after PMNN removed the internal physical equations from the model.

'mutilcontrolValueAndStabilityModelUnetRE.py'represents the model code obtained after removing all the physical equations from both the internal physical equations of the PMNN model and the loss function.

'mutilcontrolVmutilcontrolValueAndStabilityModelUnetPINNRE_ConvGRU' represents the code for replacing the stability prediction branch part of the model 'mutilcontrolVmutilcontrolValueAndStabilityModelUnetPINNRE' with the ConvGRU model.

'mutilcontrolVmutilcontrolValueAndStabilityModelUnetPINNRE_ConvLSTM' represents the code for replacing the stability prediction branch part of the model 'mutilcontrolVmutilcontrolValueAndStabilityModelUnetPINNRE' with the ConvLSTM model.

'mutilcontrolValueAndStabilityModelUnetPINNRE_FineTuning.py' represents a two-stage training strategy is employed: simulation data are first applied pre-train the network, and real monitoring data are subsequently utilized for fine-tuning, thereby enhancing the generalizability and applicability of the framework to real-world conditions.

Dataset Usage Instructions:

The compressed files in the dataset need to be fully decompressed into the directory of the dataset file.

Code experimental environment:

The code is run under Python 3.11.11 and PyTorch 2.5.1 versions. The detailed library file versions can be found in the requirements file. The recommended code can be run on NVIDIA graphics cards with video memory of at least 24GB.1