Categories
Feng Li News

New paper “Escalator accident mechanism analysis and injury prediction approaches in heavy capacity metro rail transit stations” published in Safety Science

Authors: Zhiru Wang, Yu Pang, Mingxin Gan, Martin Skitmore, Feng Li

Link: https://doi.org/10.1016/j.ssci.2022.105850

Summary:

The semi-open character with high passenger flow in Metro Rail Transport Stations (MRTS) makes safety management of human-electromechanical interaction escalator systems more complex. Safety management should not consider only single failures, but also the complex interactions in the system. This study applies task driven behavior theory and system theory to reveal a generic framework of the MRTS escalator accident mechanism and uses Lasso-Logistic Regression (LLR) for escalator injury prediction. Escalator accidents in the Beijing MRTS are used as a case study to estimate the applicability of the methodologies. The main results affirm that the application of System-Theoretical Process Analysis (STPA) and Task Driven Accident Process Analysis (TDAPA) to the generic escalator accident mechanism reveals non-failure state task driven passenger behaviors and constraints on safety that are not addressed in previous studies. The results also confirm that LLR is able to predict escalator accidents where there is a relatively large number of variables with limited observations. Additionally, increasing the amount of data improves the prediction accuracy for all three types of injuries in the case study, suggesting the LLR model has good extrapolation ability. The results can be applied in MRTS as instruments for both escalator accident investigation and accident prevention.

The combined application of the TDAPA with STPA approaches provides, for the first time, an understanding of the safety constraints and non-failure state of components and passengers to be considered in the generic formation mechanism of escalator accidents. Moreover, our analysis confirms the LLR model’s validity for use in the hazard identification and injury prediction where there is a large number of variables and relatively small number of observations. Despite the case study not providing a full appreciation of the risk factors related to escalator accidents, it demonstrates the applicability of the methodology for MRTS escalator accidents.

STPA gives a wider perspective of safety management than assessing and mitigating single hazards by component failure and operation error analysis. Focusing on independent failures in safety management can only reduce the symptoms of the underlying problem – for a top-down approach on a subset of imposing hazards, a system level of safety requires the identification of all the risk factors that may cause injury. Additionally, through the “hazard” identification in STPA, the escalator accident investigation process of TDAPA provides a more complete accident reconstruction.

On the one hand, the application of TDAPA supplements the insufficiency of non-failure passenger behaviors in the STPA approach. Although non-failure task-driven passenger behaviors are not regarded as disturbances, they can easily develop into an injury accident when interacting with disturbances or the violation of safety constraints. The quantitative results of this study demonstrate that task-driven passenger behavior is a main contribution factor for bruising. On the other hand, TDAPA transforms the analysis process of STPA into a language that is easy for escalator accident investigators to understand. The traditional methods of escalator accident investigation have limitations over critical information collection for accident reconstruction. This makes it impossible to draw meaningful conclusions for suitable countermeasures to be adopted (Chi et al., 2006Xing et al., 2019).

In contrast with metro operation companies in China, which do not use professional subway station escalator incident investigation instruments, U.S.’s Consumer Product Safety Commission (USCPSC) has a National Electronic Injury Surveillance System (NEISS) to record such incidents (NEISS, 2018). In this database, the escalator-related incidents are recorded in a simple, factual manner: for example, “30 years old intoxicated male fell 20ft over side of escalator at subway” and “46 years old female was at *** station when toe got caught”(NEISS, 2018). The results show that the five simple questions provided by the TDAPA are capable of accident reconstruction and quantitative accident analysis.

The case study results indicate that the LLR model can not only be used to identify interdependent risk factors, but also perform parameter estimation analysis – it overcomes the shortcomings of the Cramer’s, Fisher, and other correlation analysis approaches applied in escalator accident analysis (Chi et al., 2006Xing et al., 2019), which can only be used in the analysis of two variables and not in parameter estimation. Meanwhile, the LLR model overcomes the requirements of network-based approaches (as applied in Wang et al., 2020aWang et al., 2020bXing et al., 2020Bress et al., 2018) that variables should follow a strict causality relationship; therefore, it can be used when the cause of accidents may be the interaction between violation of safety constraints and the non-failure state of contribution factors as generated by TDAPA and STPA approaches in the generic formation mechanism of escalator accident analysis.

Systematic factors containing a single failure, and such potential risk factors as non-failure task-driven and task-irrelative passenger behaviors, safety constraints, and passive passenger actions were also identified by the LLR model in the case study, while such potential risk factors as task-driven passenger behaviors, safety constraints, and hazards were not identified as significant factors in the Logistic model. This is because the Logistic model requires all variables in the model to be independent, so that the multicollinearity of the full factorial Logistic regression model can lead to nonsignificant test statistics of some variables. While variable selection can avoid multicollinearity, some important variables may be removed in the process of subset selection: the results show that the LLR model deals well with multicollinearity and estimate parameters by means of regression coefficient compression, which can effectively avoid the reservation and deletion process in the selection of subset variables.

In addition, the prediction test could not be carried out on a Logistic model, because of the zero frequency of some variables. Having a relatively large number of variables and relatively small number of observations can cause a floating-point overflow error when calculating prediction accuracy by the Logistic regression model. In contrast, LLR is particularly useful when the number of variables is larger than the number of observations: as the case study result demonstrates, its prediction accuracy is higher than the Logistic model. Increasing the amount of data also improves the predictive accuracy for all three types of injuries, which suggests that the LLR model has good extrapolation properties.

Based on the analysis results, relevant safety constraints can be strengthened in system design, operation, and maintenance to prevent escalator accidents. Although there is no single failure that points to engineering design flaws, the potential hazards of some risk factors identified by TDAPA and STPA may be engineering design flaws: for example, passengers with luggage having to take the escalator as it is not convenient to take the elevator, passengers can carry large luggage on the escalator as there is no restriction device on the escalator’s entrance, and the surrounding area is crowded and passengers cannot hold onto the escalator handrail because of the lack of a flow controlling device at the escalator entrance.

For the Beijing MRTS, as the contribution level of passive interaction with other passengers is the highest of all injury-related risk factors, the safety manager should consider making improvements from the operation perspective. In China, passengers always stand directly in front of each other on the right side of the escalator, especially during MRTS peak hours, which is one of the reasons why passengers are crushed by riders ahead or behind. Accordingly, riders need to be encouraged to stand one person per step and alternately left and right, which will leave a safety buffer step and longer passenger reaction distance. Additionally, the maintenance problems identified as being associated with bruise injuries point to the need to strengthen safety management in escalator maintenance, especially inspections for water absorbing blanket bulges, exposure of damaged escalator parts, camouflage doors, and escalator floor plate bulges.

Leave a Reply

Your email address will not be published.