Seeing Structural Failure Before it Happens: An Image-Based Physics-Informed Neural Network (PINN) for Spaghetti Bridge Load Prediction

TL;DR

This study demonstrates that physics-informed neural networks can accurately predict spaghetti-bridge weights using limited data by embedding structural physics into the learning process. It introduces a novel Physics Informed Kolmogorov Arnold Network (PIKAN) alongside a PINN baseline, and couples them with an image-based parameter extraction pipeline to enable dual input modes. With an augmented dataset of 100 samples, both approaches achieve $R^2=0.9603$ and $MAE=10.50$ units, illustrating data-efficient, physically consistent predictions. A web interface provides accessible parameter entry and prediction, underscoring the approach's educational value and potential for early-stage design assessment in lightweight structures.

Abstract

Physics Informed Neural Networks (PINNs) are gaining attention for their ability to embed physical laws into deep learning models, which is particularly useful in structural engineering tasks with limited data. This paper aims to explore the use of PINNs to predict the weight of small scale spaghetti bridges, a task relevant to understanding load limits and potential failure modes in simplified structural models. Our proposed framework incorporates physics-based constraints to the prediction model for improved performance. In addition to standard PINNs, we introduce a novel architecture named Physics Informed Kolmogorov Arnold Network (PIKAN), which blends universal function approximation theory with physical insights. The structural parameters provided as input to the model are collected either manually or through computer vision methods. Our dataset includes 15 real bridges, augmented to 100 samples, and our best model achieves an $R^2$ score of 0.9603 and a mean absolute error (MAE) of 10.50 units. From applied perspective, we also provide a web based interface for parameter entry and prediction. These results show that PINNs can offer reliable estimates of structural weight, even with limited data, and may help inform early stage failure analysis in lightweight bridge designs. The complete data and code are available at https://github.com/OmerJauhar/PINNS-For-Spaghetti-Bridges.

Figures (14)

• Figure 1: Overview of the complete methodology pipeline: (1) Initial dataset of 15 real bridges with measured geometric and material properties, (2) Data augmentation through parameter variation, noise addition, and physics-consistent transformations to generate 100 samples, (3) Preprocessing including standardization and 80-20 train-test split, (4) Training of PINN and PIKAN models with physics-informed loss functions, (5) Predictions on test set, (6) Evaluation metrics achieving MAE of 10.50g, RMSE of 13.37g, and R² of 0.9603, (7) Result visualization and comparison including physics loss component analysis.
• Figure 2: Step-by-step visualization of the computer vision pipeline for structural parameter extraction from bridge images. (a) Original input image. (b) Preprocessing with grayscale conversion and Gaussian blur. (c) Edge detection using Laplacian of Gaussian. (d) Binary edge mask generation with thresholding. (e) Corner detection with FAST algorithm. (f) Corner filtering to identify structurally significant points. (g) Final angle of inclination calculation with connecting lines.
• Figure 3: Base neural network architecture used for weight prediction. The model consists of three fully connected hidden layers with ReLU activation, Batch Normalization, and Dropout regularization.
• Figure 4: Schematic of the Physics-Informed Loss Function used in the PIKAN model, showing the combination of data-driven and physics-driven loss components.
• Figure 5: Illustration of the Truncated Polynomial Layer used in the PIKAN model. The layer expands input features into higher-order polynomial terms and pairwise interactions up to a fixed degree.