Estimation of Constant Speed Time for Railway Vehicles by Stochastic Gradient Descent Algorithm

AKÇAY, Mehmet Taciddin · 2020 · Crossref

DOI: 10.35377/saucis.03.03.805598

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the optimization of railway vehicle speed profiles to improve the efficiency of rail transportation systems, specifically focusing on urban metro operations. Accurate estimation of the constant speed time—the duration a vehicle travels at its maximum operating speed between stations—is critical for optimizing energy consumption, signaling system performance, and fleet management. The research aims to estimate this parameter using machine learning techniques, proposing the Stochastic Gradient Descent (SGD) algorithm as a superior method compared to other established models. The methodology involved generating a dataset of 480 samples through simulation, varying operational conditions such as line geometry, traction force, vehicle weight, and station distances. The input variables included access distance, train resistance, vertical and lateral resistance, traction force, maximum speed, and vehicle weight, with the constant speed time serving as the output. The study utilized the Orange machine learning platform to train and evaluate seven different algorithms: Stochastic Gradient Descent, AdaBoost, Random Forest, Neural Network, k-Nearest Neighbors (kNN), Decision Trees, and Support Vector Machines (SVM). Model performance was assessed using Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the Coefficient of Determination ($R^2$). Two validation strategies were employed: 10-fold cross-validation and a random sampling hold-out method, where 70% of the data was used for training and 30% for testing. The results demonstrated that the Stochastic Gradient Descent method outperformed all other tested algorithms. Under cross-validation, SGD achieved an $R^2$ value of 0.9956, with an MSE of 6.82 and an MAE of 1.71. In the random sampling hold-out test, SGD maintained high accuracy with an $R^2$ of 0.9951, an MSE of 7.60, and an MAE of 1.76. In contrast, other methods showed lower performance; for instance, kNN and SVM yielded significantly lower $R^2$ values (below 0.90 in some cases) and higher error metrics. The study confirmed that SGD provided the most precise estimates for constant speed time, with error distributions remaining consistent across both validation methods. The significance of this work lies in demonstrating the efficacy of machine learning, particularly SGD, in predicting critical operational parameters for rail systems. By accurately estimating constant speed time, operators can optimize driving profiles, reduce energy consumption, and enhance the reliability of signaling systems. The findings support the integration of advanced machine learning algorithms into railway management systems to handle dynamic operating conditions and improve overall traffic efficiency.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• speed choice