Python Implemented AI based Machine Learning Approach for the Prediction of Optimum Location of Building using Structural Response with Self Learning using ANN

Abstract

The construction industry faces growing demands for faster, safer, and cost-effective building designs. Traditional methods are no longer sufficient, prompting the use of AI and ML for data-driven optimization. This study analyzes 10 multistory building cases with varying soil stiffness (K) using structural analysis software. A Python-based ML model focuses on predicting optimal configurations using column axial force data. An ANN is developed to minimize axial forces, with outputs visualized using Matplotlib. Data preprocessing is done using Pandas and NumPy, and models are built with scikit-learn and TensorFlow. Both Linear Regression and ANN are applied, with an 80:20 train-test split. The ANN outperforms with an MSE of 0 and R² of 1 after 150 training epochs. The model identifies optimal designs, improving cost efficiency and structural stability. This approach enhances design accuracy and reduces manual effort in structural engineering tasks

Introduction

1. Introduction to AI & ML

• Artificial Intelligence (AI): A branch of computer science focused on simulating human intelligence—learning, reasoning, and decision-making.

• Machine Learning (ML): A subset of AI that uses data-driven algorithms to predict or automate decisions without explicit programming. It recognizes patterns and learns from data.

2. Importance of Site Selection in Multi-Storey Buildings

• Load-bearing capacity and soil type are critical for foundation design.

• Extra load capacity in buildings enhances:

  • Safety margins for unpredictable loads (e.g., seismic activity).

  • Durability and lifespan of the structure.

  • Flexibility for future renovations or expansions.

  • Resilience to dynamic or impact loads.

  • Lower maintenance costs over time.

3. Literature Review Highlights

• Pandelea et al. (2014): AI (especially Artificial Neural Networks) applied in various civil engineering domains—materials, geotech, and structural behavior.

• Majumdar et al. (2016): Use of ML in developing efficient sorting algorithms using genetic and algebraic systems.

• Zhang et al. (2021): Robotic sorting system in unstructured scenes using Mask R-CNN and ML for accurate object detection and grasping.

• Purizaca et al. (2020): Review of 41 papers showing ANN’s effectiveness in predicting civil engineering variables with high accuracy (R² ≈ 0.99).

• Patil et al. (2021): ML-based grading and sorting of dragon fruit using CNN, ANN, and SVM for sustainable agriculture.

4. Identified Research Gaps

• Lack of integrated frameworks combining structural response data and optimization.

• Most models use synthetic data, lacking real-world applicability.

• Generalized algorithms don’t account for unique building designs.

• Very few efforts exist for Python-based AI applications in structural design.

• Limited use of multi-variable optimization and adaptive/self-learning models.

5. Research Objectives

• Develop building models using structural software across various borehole soil conditions.

• Extract spring stiffness and axial force data for AI training.

• Build an AI/ML model in Python for:

  • Data cleaning and normalization.

  • Model training and validation.

  • Predicting optimal, cost-effective structural designs.

• Evaluate model performance using metrics like MSE and R² score.

6. Methodology Overview

Data Collection & Preprocessing

• Sources: Public datasets and domain-specific geotechnical reports (Zone III soil).

• Methods: Z-score normalization, PCA, missing value handling.

Key Features Used

• Axial force: Represents longitudinal internal force (compression/tension).

• Horizontal spring stiffness: Measures lateral resistance of foundation (K = P/?).

7. ML Workflow (Python Example with Linear Regression)

• Step 1: Clean data and set up input (X) and target (y) variables.

• Step 2: Split data into training/testing sets.

• Step 3: Train regression model.

• Step 4: Make predictions.

• Step 5: Evaluate using Mean Squared Error (MSE) and R² score.

8. Output and Analysis

• Several building models were created and analyzed based on borehole soil types.

• Maximum axial forces were compared.

• Flowcharts, tables, and 3D views demonstrate the model-building and ML stages.

• Result: AI/ML models can effectively predict optimal structural configurations for different soil conditions.

Conclusion

With the aforementioned problem, the following conclusions have drawn mentioned below: 1) Building case creation: Multiple structural models were successfully created using building analysis software, incorporating varied subsurface conditions represented by different borehole data. This enabled a realistic evaluation of the influence of geotechnical variability on structural performance. Total 10 distinct building cases created abbreviated as OBM1 to OBM10. 2) Spring stiffness calculation:Spring stiffness values were accurately derived from the soil investigation reports, allowing realistic simulation of soil structure interaction and the incorporation of subgrade response in the structural models. It has basically denoted by K and the obtained value used as input for analysis software to provide a pretend scenario over actual soil. 3) Axial force analysis:Structural analysis yielded axial force values as key output parameters. These results formed the basis for evaluating the structural loading behavior under varying soil conditions and served as critical input data for AI-ML model training. 4) AI-ML based code development:A Python based machine learning model was developed incorporating essential pre processing steps such as missing value treatment, data normalization and data cleaning to ensure high quality and reliable input data. This has achieved using Pandas v2.0.3 and NumPy v1.26.4. 5) Selection of techniques for data relationship:The AI-ML based linear regression and feed forward Artificial Neural Network model established using meaningful relationships between the input parameters (e.g., soil properties, stiffness, load cases) and output responses (axial force), demonstrating strong predictive capability. 6) Validation technique selection:Suitable validation techniques such as scikit-learn (sklearn) v1.3.0 and TensorFlow v2.16.1 used that train test split data and conduction of cross validation to ensure robust evaluation of model performance. An 80:20 train test split was applied to the dataset, and the models were trained and validated using scikit-learn (sklearn) v1.3.0 and TensorFlow v2.16.1 respectively. 7) Model evaluation (MSE and R²):Model performance was assessed using Mean Squared Error (MSE) and R² score. The results indicate that ANN trained over 150 epochs (time) with two hidden layers of 64 and 32 neurons, produced an MSE of 0 and an R² of 1, thereby demonstrating improved predictive accuracy and reliability. 8) Optimum case comparison:Among all structural cases, the model successfully identified the optimum configuration i.e., the case with the least axial force highlighting the effectiveness of AI-ML in optimizing structural design based on soil and structural parameters. The main and foremost objective has achieved, all the model cases have created over selected bore hole cases, applied K value, analyzed and results obtained by using analysis software. Results are then used as input for obtaining the best possible location. Also, the AI and ML based code snippets has proved for fast building site prediction with larger dataset with recommendation of economic case among different soil scenario and will recommend to use this self learning approach for practical application in building engineering which significantly minimized manual effort while improving the speed and accuracy of structural assessments.

References

[1] Pandelea, A.-E. C., Budescu, M. Gh., Covatariu, G. M., & Gheorghe Asachi Technical University of Ia?i, Faculty of Civil Engineering and Building Services. (2014). Applications of artificial neural networks in civil engineering. In 2nd International Conference for PhD Students in Civil Engineering and Architecture. [2] Sasnauskas, R., SES Engineering, Chen, Y., Nvidia, Inc., Collingbourne, P., Google, Inc., Ketema, J., Embedded Systems Innovation by TNO, Lup, G., Microsoft, Inc., Taneja, J., University of Utah, Regehr, J., & University of Utah. (n.d.). A Synthesizing Superoptimizer. Conference Paper. [3] Shazeer, N. & Google. (2019). Fast Transformer Decoding: One Write-Head is All You Need. Conference Paper. [4] Evaluating Large Language Models Trained on Code. (2021). In arXiv [Journal-article]. https://arxiv.org/pdf/2107.03374 [5] Pearce, H., Tan, B., Ahmad, B., Karri, R., New York University, & University of Calgary. (2022). Examining Zero-Shot Vulnerability Repair with Large Language Models [Journal-article]. arXiv. https://arxiv.org/abs/2112.02125v3 [6] Andon, A., &Covatariu, G. (2022). A Study on Image Processing Using Artificial Neural Networks in Civil Engineering. BuletinulInstitutuluiPolitehnic “Gheorghe Asachi” Din Ia?i. Sec?iaConstruc?ii. Arhitectur?, 67(3), 85–94. https://doi.org/10.2478/bipca-2021-0027 [7] Majumdar, S., Jain, I., & Kukreja, K. (2016). AdaSort: Adaptive Sorting using Machine Learning. International Journal of Computer Applications, 145(12), 12–17. https://doi.org/10.5120/ijca2016910726 [8] Shaikh, A., Sawant, S., Patil, A., Bhandare, A., & International Research Journal of Modernization in Engineering, Technology and Science. (2024). RAISIN SORTING USING MACHINE LEARNING ALGORITHM. International Research Journal of Modernization in Engineering Technology and Science, 06. https://www.irjmets.com [9] Fuchsgruber, D., Po?stuvan, T., G ¨Unnemann, S., Geisler, S., Department of Computer Science & Munich Data Science Institute, TU Munich, & EPFL. (2024). GRAPH NEURAL NETWORKS FOR EDGE SIGNALS: ORIENTATION EQUIVARIANCE AND INVARIANCE. [10] Zhang, H., Liang, H., Ni, T., Huang, L., & Yang, J. (2021). Research on Multi-Object Sorting System Based on Deep Learning. Sensors, 21(18), 6238. https://doi.org/10.3390/s21186238 [11] Purizaca, F. J. V., Barturen, D. a. C., Pérez, S. P. M., Tuesta-Monteza, V. A., & Mejía-Cabrera, H. I. (2020). Importance of artificial neural networks in civil engineering: a systematic review of the literature. ITECKNE Innovación E Investigación En Ingeniería, 18(1). https://doi.org/10.15332/iteckne.v18i1.2542 [12] Anjum, A., Hrairi, M., Shaikh, A. A., Yatim, N., & Ali, M. (2024). Integrating AI and statistical methods for enhancing civil structural practices: current trends, practical issues, and future direction. Frattura Ed IntegritàStrutturale, 19(71), 164–181. https://doi.org/10.3221/igf-esis.71.12 [13] Li, X., Garzaran, M. J., & Padua, D. (2007). Optimizing Sorting with Machine Learning Algorithms. Conference Paper, 1 2, 1–6. https://doi.org/10.1109/ipdps.2007.370499 [14] Patil, P. U., Lande, S. B., Nagalkar, V. J., Nikam, S. B., & Wakchaure, G. (2021). Grading and sorting technique of dragon fruits using machine learning algorithms. Journal of Agriculture and Food Research, 4, 100118. https://doi.org/10.1016/j.jafr.2021.100118 [15] Research and applications of artificial neural network in pavement engineering: A state-of-the-art review. (2021). In Journal of Traffic and Transportation Engineering (English Edition). http://doi.org/10.1016/j.jtte.2021.03.005 [16] Faster sorting algorithms discovered using deep reinforcement learning. (2023). Nature, 618, 257–258.

Copyright

Copyright © 2025 Shakila Begam, Ms. Ashu Nayak. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.