Domain-Separated Quantum Neural Network for Truss Structural Analysis with Mechanics-Informed Constraints.

Area of Science:

• Quantum Computing
• Structural Engineering
• Artificial Intelligence

Background:

• Traditional structural analysis methods can be computationally intensive.
• Coordinate-based neural network models face limitations in flexibility and scalability for complex structures.

Purpose of the Study:

• To propose an index-based quantum neural network (QNN) model for static analysis of truss structures.
• To enhance the flexibility, scalability, and efficiency of structural analysis using quantum computing principles.

Main Methods:

• Developed an index-based quantum neural network (QNN) model utilizing a variational quantum circuit (VQC).
• Implemented a separate-domain strategy for parallel training of independent quantum circuits assigned to structural partitions.
• Formulated a mechanics-informed loss function based on the force method within a Lagrangian dual framework to enforce physical constraints.

Main Results:

• The QNN model achieved high prediction accuracy and fast convergence, even for complex structural conditions.
• Reduced the number of parameters by up to 64% compared to conventional neural networks while improving accuracy.
• The separate-domain approach within the QNN architecture showed a 6.25% parameter reduction over single-domain models.

Conclusions:

• The index-based QNN model demonstrates practical applicability and superior performance in static structural analysis.
• This quantum-based approach offers a powerful and efficient numerical analysis tool with potential for structural optimization and broader engineering applications.