A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Long-term prediction for railway track geometry based on an optimised DNN method
https://orcid.org/0000-0003-3131-8217
https://orcid.org/0000-0003-0354-6167
Highlights The railway track longitudinal levels collected at high measurement frequency (monthly) in the Netherlands are analyzed. The optimised DNN model using the Random Search algorithm is proposed. The short-term and long-term prediction accuracy of the optimised DNN model is both high while the customization time is shortened. The growth pattern and the location characteristics of track geometry can be predicted and the maintenance effect can be considered by the optimised DNN model. The prediction results can advise the time and locations for track maintenance.
Abstract Railway transportation becomes increasingly important due to the rapidly emerging needs of global trade, environmental protection, and high oil costs. Thus, railway infrastructure companies are required to provide higher reliability and safety for railway transportation. With the fast development of computer science, Deep Neural Network (DNN) models are developed for track geometry prediction to improve the efficiency of track maintenance. However, existing DNN models are still not widely adopted by railway companies because they are often needed to be re-programmed intensively by experienced engineers for customization. Therefore, an optimised DNN model using the Random Search algorithm is developed in this paper for track geometry prediction. The track longitudinal level measured for 2 years on a monthly basis is used to train and validate the prediction model. The results show short-term and long-term prediction accuracy of the optimised DNN model are both high while the customization time is shortened. The growth pattern and the location characteristics of track geometry can be predicted and the maintenance effect can be considered. The prediction results can advise the time and locations for track maintenance.
Long-term prediction for railway track geometry based on an optimised DNN method
https://orcid.org/0000-0003-3131-8217
https://orcid.org/0000-0003-0354-6167
Highlights The railway track longitudinal levels collected at high measurement frequency (monthly) in the Netherlands are analyzed. The optimised DNN model using the Random Search algorithm is proposed. The short-term and long-term prediction accuracy of the optimised DNN model is both high while the customization time is shortened. The growth pattern and the location characteristics of track geometry can be predicted and the maintenance effect can be considered by the optimised DNN model. The prediction results can advise the time and locations for track maintenance.
Abstract Railway transportation becomes increasingly important due to the rapidly emerging needs of global trade, environmental protection, and high oil costs. Thus, railway infrastructure companies are required to provide higher reliability and safety for railway transportation. With the fast development of computer science, Deep Neural Network (DNN) models are developed for track geometry prediction to improve the efficiency of track maintenance. However, existing DNN models are still not widely adopted by railway companies because they are often needed to be re-programmed intensively by experienced engineers for customization. Therefore, an optimised DNN model using the Random Search algorithm is developed in this paper for track geometry prediction. The track longitudinal level measured for 2 years on a monthly basis is used to train and validate the prediction model. The results show short-term and long-term prediction accuracy of the optimised DNN model are both high while the customization time is shortened. The growth pattern and the location characteristics of track geometry can be predicted and the maintenance effect can be considered. The prediction results can advise the time and locations for track maintenance.
Long-term prediction for railway track geometry based on an optimised DNN method
https://orcid.org/0000-0003-3131-8217
https://orcid.org/0000-0003-0354-6167
Highlights The railway track longitudinal levels collected at high measurement frequency (monthly) in the Netherlands are analyzed. The optimised DNN model using the Random Search algorithm is proposed. The short-term and long-term prediction accuracy of the optimised DNN model is both high while the customization time is shortened. The growth pattern and the location characteristics of track geometry can be predicted and the maintenance effect can be considered by the optimised DNN model. The prediction results can advise the time and locations for track maintenance.
Abstract Railway transportation becomes increasingly important due to the rapidly emerging needs of global trade, environmental protection, and high oil costs. Thus, railway infrastructure companies are required to provide higher reliability and safety for railway transportation. With the fast development of computer science, Deep Neural Network (DNN) models are developed for track geometry prediction to improve the efficiency of track maintenance. However, existing DNN models are still not widely adopted by railway companies because they are often needed to be re-programmed intensively by experienced engineers for customization. Therefore, an optimised DNN model using the Random Search algorithm is developed in this paper for track geometry prediction. The track longitudinal level measured for 2 years on a monthly basis is used to train and validate the prediction model. The results show short-term and long-term prediction accuracy of the optimised DNN model are both high while the customization time is shortened. The growth pattern and the location characteristics of track geometry can be predicted and the maintenance effect can be considered. The prediction results can advise the time and locations for track maintenance.
Long-term prediction for railway track geometry based on an optimised DNN method
Han, Lei (author) / Liao, Yingying (author) / Wang, Haoyu (author) / Zhang, Hougui (author)
2023-07-30
Article (Journal)
Electronic Resource
English
Prediction of railway track geometry defects: a case study
| Taylor & Francis Verlag | 2020
Modelling railway track geometry deterioration
| Online Contents | 2011
Analytical forecasting of long-term railway track settlement
| Elsevier | 2021