Table 4 Top 10 most relevant articles in... | Emerald Publishing

Table 4

Top 10 most relevant articles in rail

Source	Objectives	AI models	Maintenance aspect	Short outcomes
Kaewunruen et al. (2021)	To optimize long-term rail renewal and maintenance schedules using deep reinforcement learning	DRL	Renewal and maintenance planning	The DRL approach generated cost-effective maintenance and renewal plans that reduced the risk of failures, showing promise in automating complex planning decisions for rail infrastructure upkeep
Famurewa et al. (2017)	Develop diagnostic, predictive and prescriptive analytics for railway infrastructure maintenance	Data analytics framework	Maintenance decision support	Showcased an analytics approach that diagnoses faults, predicts failures and prescribes maintenance actions, improving decision-making for infrastructure upkeep and reliability
Khouzani et al. (2017)	Model track geometry degradation stochastically for optimized maintenance scheduling	Stochastic degradation modeling	Maintenance planning	The model forecasts track geometry deterioration over time, helping planners schedule maintenance (e.g. tamping) at optimal times to balance cost and safety
Gao et al. (2018)	Integrate multiple rail inspection data sources to detect surface and sub-surface defects	DF and ML	Predictive maintenance	Improved defect detection accuracy by combining diverse data (e.g. ultrasonic, visual inspections), enabling earlier identification of rail flaws and reducing the risk of failures
Gerum et al. (2019)	To develop a data-driven policy for scheduling railway maintenance activities predictively	ML	Maintenance scheduling	Demonstrated that using data-driven models to schedule track maintenance can effectively prevent failures (like track buckling or breaks) and optimize maintenance intervals
Hu and Liu (2016)	To predict railway track geometry degradation using machine learning	SVM	Preventive maintenance scheduling	The SVM model accurately predicted track geometry degradation trends, aiding proactive planning of maintenance before track quality falls below safety thresholds
Zhang et al. (2020)	Predict the occurrence of broken rails using machine learning techniques	ML classification (e.g. random forest)	Maintenance planning	Achieved the ability to identify high-risk track segments for rail breaks, allowing preventive repairs to be scheduled and significantly enhancing railway safety
Li et al. (2014)	Analyze factors affecting train delays and network speed using machine learning	ML	Maintenance scheduling and planning	Identified maintenance-related factors that impact network velocity (on-time performance), informing adjustments to maintenance schedules to reduce delays and improve overall service punctuality
Lasisi and Attoh-Okine (2019)	Enhance the accuracy of rail defect predictions using ensemble learning methods	EML	Condition-based maintenance scheduling	Ensemble models improved the precision of defect detection and prediction, leading to more targeted maintenance actions and more efficient use of repair resources on the railway network
Mohammadi and He (2022)	Monitor rail corrugation and predict track wear to extend the service life of tracks and rolling stock	ML	Maintenance planning	Enabled early detection of rail corrugation issues, allowing timely grinding maintenance and prolonging the lifespan of track and wheel assets

Source	Objectives	AI models	Maintenance aspect	Short outcomes
Kaewunruen et al. (2021)	To optimize long-term rail renewal and maintenance schedules using deep reinforcement learning	DRL	Renewal and maintenance planning	The DRL approach generated cost-effective maintenance and renewal plans that reduced the risk of failures, showing promise in automating complex planning decisions for rail infrastructure upkeep
Famurewa et al. (2017)	Develop diagnostic, predictive and prescriptive analytics for railway infrastructure maintenance	Data analytics framework	Maintenance decision support	Showcased an analytics approach that diagnoses faults, predicts failures and prescribes maintenance actions, improving decision-making for infrastructure upkeep and reliability
Khouzani et al. (2017)	Model track geometry degradation stochastically for optimized maintenance scheduling	Stochastic degradation modeling	Maintenance planning	The model forecasts track geometry deterioration over time, helping planners schedule maintenance (e.g. tamping) at optimal times to balance cost and safety
Gao et al. (2018)	Integrate multiple rail inspection data sources to detect surface and sub-surface defects	DF and ML	Predictive maintenance	Improved defect detection accuracy by combining diverse data (e.g. ultrasonic, visual inspections), enabling earlier identification of rail flaws and reducing the risk of failures
Gerum et al. (2019)	To develop a data-driven policy for scheduling railway maintenance activities predictively	ML	Maintenance scheduling	Demonstrated that using data-driven models to schedule track maintenance can effectively prevent failures (like track buckling or breaks) and optimize maintenance intervals
Hu and Liu (2016)	To predict railway track geometry degradation using machine learning	SVM	Preventive maintenance scheduling	The SVM model accurately predicted track geometry degradation trends, aiding proactive planning of maintenance before track quality falls below safety thresholds
Zhang et al. (2020)	Predict the occurrence of broken rails using machine learning techniques	ML classification (e.g. random forest)	Maintenance planning	Achieved the ability to identify high-risk track segments for rail breaks, allowing preventive repairs to be scheduled and significantly enhancing railway safety
Li et al. (2014)	Analyze factors affecting train delays and network speed using machine learning	ML	Maintenance scheduling and planning	Identified maintenance-related factors that impact network velocity (on-time performance), informing adjustments to maintenance schedules to reduce delays and improve overall service punctuality
Lasisi and Attoh-Okine (2019)	Enhance the accuracy of rail defect predictions using ensemble learning methods	EML	Condition-based maintenance scheduling	Ensemble models improved the precision of defect detection and prediction, leading to more targeted maintenance actions and more efficient use of repair resources on the railway network
Mohammadi and He (2022)	Monitor rail corrugation and predict track wear to extend the service life of tracks and rolling stock	ML	Maintenance planning	Enabled early detection of rail corrugation issues, allowing timely grinding maintenance and prolonging the lifespan of track and wheel assets

Note(s): ML: Machine learning; DF: data fusion; SVM: support vector machine; EML: ensemble machine learning, DRL: deep reinforcement learning