HARDING PRIZE COMPETITION 2023 | BTS
By utilising the latest research, a successful industry-
driven study was delivered. The Southwark Tunnel model was able to predict
the EPB, ring cycle duration, and rings per shift from the same neural network. The algorithm was efficient, yet still accurate, and is valuable for higher-level deployment. Predicting machine parameters, such as EPB, is
useful for autonomous projects, or as a driver-assist to recommend to an operator what controls they can consider for adjustment. The algorithm could also extract an innate understanding of the data without being heavily influenced by the operational data, as shown by R2
values being lower than 0.9. TBM data
quality is important, as performance depends on data quality. The NBRS model was shown to successfully forecast
ring cycle time, a key metric for anyone planning or delivering a tunnelling project. It was also able to predict spikes and deviations, despite having no prior knowledge of the ground conditions. The conditions that it lacked at training were
sandstone and gypsum, but the neural network was able to correctly anticipate their impact. The approach that this neural network used
was novel as it didn’t use TBM performance data as inputs, meaning it could be used at the tendering stage.
There is opportunity to further improve the model by
including the tunnel geometry, which may influence the performance. This can also be analysed to look for relationships between the geometry of the tunnel, ground condition and the time it takes to construct one ring. Data of unique ground conditions can be introduced, such as shear, moisture, and particle size. This would help enable easier training of a model, as it can find a relationship between the time to excavate and the ground conditions, with respect to other ground conditions. The neural network could also be adapted for real-
time project updates. By taking in machine parameters in certain conditions, projections can be adjusted. Collecting more of this data would also be integral for any future work that considers using data-driven technology to improve performance. By formulating a digital technology roadmap, with standardised, quality data at its core, multiple models for different tasks can be developed across the business, and the industry as a whole. To conclude, this study demonstrated the successful
application of artificial intelligence for tunnelling production forecasting. This was a novel use of a new technology that helped improve efficiency by understanding productivity by using a data-driven approach. Murphy is now taking further steps in this area to develop the work.
REFERENCES 1 Goodfellow, I., Bengio, Y., and Courville, A. (2016) Deep Learning. MIT Press. Accessed: Jan. 31, 2023. [Online]. Available:
https://www.deeplearningbook.org/
2 Borg, M. et al., (2018) Safely Entering the Deep: A Review of Verification and Validation for Machine Learning and a Challenge Elicitation in the Automotive Industry. Accessed: Jan. 31, 2023. [Online]. Available:
https://medium.com/@karpathy/software- 2-0-a64152b37c35
3 Shailaja, K., Seetharamulu, B. and Jabbar, M.A. (2018) Machine Learning in Healthcare: A Review, Proceedings of the 2nd International Conference on Electronics, Communication and Aerospace Technology, ICECA 2018 pp. 910–914, Sep. 2018, doi: 10.1109/ICECA.2018.8474918
4 Dixon, M.F., Halperin, I. and Bilokon, P. (2020) Machine learning in finance: From theory to practice. Machine Learning in Finance: From Theory to Practice, pp. 1–548, Jan. 2020, doi: 10.1007/978-3-030-41068-1/COVER
5 Xu, Y., Zhou, Y., Sekula, P. and Ding, L. (2021) Machine learning in construction: From shallow to deep learning. Developments in the Built Environment, vol. 6, p. 100045, May 2021, doi: 10.1016/J.DIBE.2021.100045
6 Liu, Y., Wang, Y. and Li, X. (2019) Computer Vision Technologies and Machine Learning Algorithms for Construction Safety Management: A Critical Review. ICCREM 2019: Innovative Construction Project Management and Construction Industrialization - Proceedings of the International Conference on Construction and Real Estate Management 2019, pp. 67–81, 2019, doi: 10.1061/9780784482308.008
7 Ayawah, P. E. A. et al., (2022) A review and case study of Artificial intelligence and Machine learning methods used for ground condition prediction ahead of tunnel boring Machines. Tunnelling and Underground Space Technology, vol. 125, p. 104497, Jul. 2022, doi: 10.1016/J.TUST.2022.104497
8 Xu, C., Liu, X., Wang, E. and Wang, S. (2021) Prediction of tunnel boring machine operating parameters using various machine learning algorithms. Tunnelling and Underground Space Technology, vol. 109, p. 103699, Mar. 2021, doi: 10.1016/J. TUST.2020.103699
9 Karpatne, A., Ebert-Uphoff, I., Ravela, S., Babaie, H. A. and Kumar, V. (2019) Machine Learning for the Geosciences: Challenges and Opportunities. IEEE Trans Knowl Data Eng, vol. 31, no. 8, pp. 1544–1554, Aug. 2019, doi: 10.1109/TKDE.2018.2861006
10 Bhatia, V. (2020) Activation Functions-How the Neuron Triggers.
https://medium.com/analytics-vidhya/activation-functions- how-should-the-neurons-trigger-1349a383ffeb (accessed Jan. 31, 2023)
August 2023 | 41
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49
