An Intelligent Evaluation System for Predicting Engine Driver Reliability
- PDF / 493,600 Bytes
- 5 Pages / 612 x 792 pts (letter) Page_size
- 87 Downloads / 225 Views
ntelligent Evaluation System for Predicting Engine Driver Reliability V. G. Sidorenkoa, * and M. A. Kulagina aRussian
University of Transport, Moscow, 127994 Russia *e-mail: [email protected]
Received June 14, 2020; revised July 22, 2020; accepted July 27, 2020
Abstract—An intelligent system is considered for evaluating and predicting engine driver reliability is considered. This is based on predicting possible engine driver’s violations, depending on his previous experience and the formation of appropriate recommendations. The problem of predicting possible violations is solved using tools and machine learning algorithms. The proposed system makes it possible to predict violations that may be committed by a driver and to increase the reliability of electric stock safety systems by targeted management of the human factor. Keywords: electric stock, driver’s performance, machine learning, recommendation algorithms, neural network, optimization DOI: 10.3103/S1068371220090126
Machine learning methods are widely used in various fields of science, as they provide efficient and quick analysis of large data arrays, predicting the onset of an event, and detecting and recognizing patterns and speech. The topic under study is related to the safety of railway transport, the provision of which is one of the key tasks of Russian Railways. Among the main factors affecting traffic safety (Fig. 1), the human factor is one of the most important; therefore, the development of methods aimed at evaluation and prediction of risks associated with errors or disruptions in the work of an engine driver is a very urgent task. The article discusses an intelligent system designed to evaluate and predict the engine driver reliability. The solution to this problem is based on predicting possible violations on the part of a driver, depending on his previous experience, the violations that have taken place, and provision of appropriate recommendations [1–3]. Engine drivers operating on electric stock (ES) commit a large number of a lot of violations that affect both the state of the ES and traffic safety (Fig. 2). For example, approximately 35% of driver violations in 2020 are related to brake control. To develop a unified and objective way of evaluating the probability of a violation and the provision of recommendations depending on the previously committed violations, it is necessary to use a mathematical
tool (mathematical models, methods, and algorithms) (Fig. 3). One way to solve this problem is to calculate the probability of a driver’s violation on his next trip, as well as the driver’s reliability level for the next 30 days. Another area includes a mathematical model for providing recommendations depending on previously committed violations. The level of risk is determined using a decision rule or an algorithm integrator. As part of the study of violations committed by drivers, it was revealed that the appearance of violations entails the appearance of similar ones. Graphic illustrations clearly demonstrate this (Fig. 4).
Data Loading...
