Real-world Accident Data Research Articles

Causal V: A five-phase causal mining methodology for real-world accident data

Causal V: A five-phase causal mining methodology for real-world accident data

Advanced analytics for predicting traffic collision severity assessment

Accurate prediction of accident risks plays a crucial role in proactively implementing safety measures and allocating resources effectively. This paper introduces an innovative approach aimed at improving accident risk prediction by harnessing unique data sources and extracting insights from diverse yet sparse datasets. Traditional models often face limitations due to a lack of diversity and scope in the available data, which hinders their predictive capabilities. In response to this challenge, our study integrates a broad spectrum of heterogeneous data, encompassing traffic flow, weather conditions, road infrastructure details, and historical accident records. To overcome the difficulties associated with sparse data, we employ advanced data science techniques such as feature engineering, imputation, and machine learning. The paper introduces a novel dataset that amalgamates various data types, establishing a robust foundation for our predictive model. Through meticulous analysis, we derive valuable insights from these diverse sources, significantly enhancing our ability to assess accident risks. The proposed approach offers numerous advantages, including the capacity to predict accidents in areas that were previously underrepresented and under varying conditions. We rigorously evaluate the model's performance through extensive experimentation and validate its accuracy using real-world accident data. Our results indicate substantial improvements in prediction accuracy compared to conventional models. This research contributes significantly to the field of accident risk prediction by highlighting the potential benefits of integrating heterogeneous sparse data and leveraging advanced data science techniques. The study underscores the importance of tapping into novel data sources and extracting concealed patterns and insights to promote safety and optimize resource allocation in accident-prone regions, fostering more secure environments.

Enhancing Accident Risk Prediction with Novel Data and Findings from Heterogeneous Sparse Sources

Accurate accident risk prediction is paramount for proactive safety measures and resource allocation. This paper introduces an innovative approach to enhance accident risk prediction by leveraging novel data sources and uncovering insights from heterogeneous sparse data. Traditional models often suffer from limitations in data diversity and scope, hindering their predictive capabilities. In response, our study integrates a wide range of heterogeneous data, including traffic flow data, weather conditions, road infrastructure, and historical accident records. To address the challenges of working with sparse data, we employ advanced data science techniques, including feature engineering, imputation, and machine learning. The paper presents a new dataset that combines diverse data types, providing a comprehensive foundation for our predictive model. Through rigorous analysis, we extract valuable insights from these heterogeneous sources to improve accident risk assessment. The proposed approach offers several advantages, including the ability to predict accidents in previously underrepresented areas and under varying conditions. We evaluate the model's performance through extensive experimentation and validate its accuracy against real-world accident data. Our findings demonstrate significant enhancements in prediction accuracy compared to conventional models. This research contributes to the field of accident risk prediction by showcasing the potential of heterogeneous sparse data integration and advanced data science techniques. It highlights the importance of utilizing novel data sources and the value of uncovering hidden patterns and insights to foster safer environments and more efficient resource allocation in accident-prone areas.

Evaluation of Brain Rotational Injury Criteria (BrIC) in vehicle frontal crashes

Objective The objective of this study was to estimate strains in the human brain in regulatory, research, and due care frontal crashes by simulating those impacts. In addition, brain strain simulations were estimated for belted human volunteer tests and in impacts between two players in National Football League (NFL), some with no injury and some with mild Traumatic Brain Injuries (mTBI). Methods The brain strain responses were determined using version 5 of the Global Human Body Modeling Consortium (GHBMC) 50th percentile human brain model. One hundred and sixty simulations with the brain model were conducted using rotational velocities and accelerations of Anthropomorphic Test Devices (ATD’s) or those of human volunteers in sled or crash tests, as inputs to the model and strain related responses like Maximum Principal Strains (MPS) and Cumulative Strain Damage Measure (CSDM) in various regions of the brain were monitored. The simulated vehicle tests ranged from sled tests at 24 and 32 kph delta-V with three-point belts without airbags to full scale crash and sled tests at 56 kph and a series of Research Mobile Deformable Barrier (RMDB) tests described in Prasad et al. Results The severity of rotational input into the model as represented by BrIC, averaged between 0.5 and 1.2 for the various test conditions, and as high as 1.5 for an individual case. The MPS responses for the various test conditions averaged between 0.28 and 0.86 and as high as 1.3 in one test condition. The MPS responses in the brain for volunteers, low velocity sled, and NCAP tests were similar to those in the no-mTBI group in the NFL cases and consistent with real world accident data. The MPS responses of the brain in angular crash and sled tests were similar to those in the mTBI group. Conclusions The brain strain estimations do not indicate the likelihood of severe-to-fatal brain injuries in the crash environments studied in this paper. However, using the risk functions associated with BrIC, severe-to-fatal brain injuries (AIS4+) are predicted in several environments in which they are not observed or expected.

Leveraging the GIDAS Database for the Criticality Analysis of Automated Driving Systems

A safe introduction of automated driving systems on urban roads requires a thorough understanding of the traffic conflicts and accidents. This understanding is paramount to constructively safeguard these systems, i.e., to design a system that exhibits an adequate performance even in critical situations. In this work, we present an approach to gather knowledge by analyzing the German In-Depth Accident Study (GIDAS) database, which is representative of all German traffic accidents, along with the influencing factors that are hypothesized to be associated with increased criticality in relation to automated driving. In order to gain an insight into the risk associated with these factors in real-world accidents, we determine their presence in the database’s accident cases within a selected operational domain, enabled by translation from a natural language description to the database scheme employed by GIDAS. This initial catalog as well as the subsequent statistical considerations is motivated by analyzing the criticality for automated driving systems in urban areas. Based on this catalog, our work delineates a method for quantification of risk associated with such influencing factors in a given operational domain based on real-world accident data. This quantification can subsequently be used in decompositional, scenario-based risk assessment before system design and for the embedding safety argumentation. This paper, therefore, provides a blueprint of how the matured field of traffic accident research studies and its results, in particular accident databases, can be leveraged for risk assessment of the operational domain of automated driving systems.

A Comparative Study Between Different NCAPs and The Future of BNVSAP

India is the third largest road network which is used by one percent of the world’s total vehicle population. On this road network six percent of the total of global road traffic accidents occur. India is the sixth-largest car market in the world but is still the only country amongst the global top 10 car markets to not have a testing program. To reduce the number of road traffic accidents, many countries have come up with safety programmes such as Euro NCAP (New Car Assessment Program), US NCAP, Latin NCAP, etc. In 2014, Indian government proposed a similar safety assessment program called as Bharat New Vehicle Safety Assessment Program (BNVSAP) but it is since delayed in its implementation due to a number of reasons. The aim of this study is to compare existing NCAPS and the new BNVSAP and propose some improvements in it by studying the experiences gained by the older NCAPS and observing the difference in the situation of road traffic in India and in other countries by analysing real world accident data.

Investigation of Injury patterns in Heavy-duty Single Vehicle crashes based on real-world accident data in Tamilnadu, India

According to the reports of NITI (National Institute of Transforming India) Aayog Freight 2018, Road freight is the prime mode (59%) of transport in India with the highest per ton-mile cost than rail or water freight (NITI Aayog, 2018). This road freight usually uses heavy-duty vehicles to transmit voluminous goods and services to the destination in time. Due to this, the heavy-duty vehicle population increased on the Indian roads. Heavy-duty vehicle crashes cause a substantial economic burden to the nation and result in more severity to the involved because of differences in weight, speed, and size. Among heavy-duty vehicle crashes, a significant proportion of crashes are heavy-duty single-vehicle crashes. Single-vehicle crashes are those crashes where the vehicle drivers either involve in self-skidding or hit a stationary object (like a tree). The purpose of this study is to investigate the injury pattern in heavy-duty single vehicle crashes. For this study, the data is extracted from the RADMS (Road Accident Database Management System) database and linked with hospital data. This data includes demographic information, road, environmental and injury characteristics. Later, descriptive statistics performed on the dataset to analyse all heavy-duty single-vehicle crashes between January 2013 and December 2018. Overall, 4704 single heavy-duty vehicle crashes occurred during this period, among which 1244 were fatal crashes. Results show that male drivers aged 26 to 64 years old suffered more fatalities (88%), followed by the 18-25 age group (8%). Examination of injury information found that heavy-duty vehicle drivers mostly sustained multiple injuries (9.05%), head injuries (5.05%), followed by leg injuries (4.29%). The results showed that specific road and environmental factors increase the chance of fatal crashes among heavy-duty vehicle drivers. Furthermore, the proposed study gives insight into the injury characteristics and key contributing factors causing heavy-duty single-vehicle crashes. Finally, this study provides appropriate countermeasures and techniques that can mitigate heavy-duty single-vehicle collisions.

Using Generalized Linear Mixed Models to Predict the Number of Roadway Accidents: A Case Study in Hamilton County, Tennessee

Introduction: A method for identifying significant predictors of roadway accident counts has been presented. This process is applied to real-world accident data collected from roadways in Hamilton County, TN. Methods: In preprocessing, an aggregation procedure based on segmenting roadways into fixed lengths has been introduced, and then accident counts within each segment have been observed according to predefined weather conditions. Based on the physical roadway characteristics associated with each individual accident record, a collection of roadway features is assigned to each segment. A mixed-effects Negative Binomial regression form is assumed to approximate the relationship between accident counts and several explanatory variables including roadway characteristics, weather conditions, and several interactions between them. Standard diagnostics and a validation procedure show that our model form is properly specified and suitably fits the data. Results: Interpreting interaction terms leads to the follow findings: 1) rural roads with cloudy conditions are associated with relative increases in accident frequency; 2) lower/moderate AADT and rainy weather are associated with relative decreases in accident frequency, while high AADT and rain are associated with relative increases in accident frequency; 3) higher AADT and wider pavements are associated with relative increases in accident frequency; and 4) higher speed limits in residential areas are associated with relative increases in accident frequency. Conclusion: Results illustrate the complicated relationship between accident frequency and both roadway features and weather. Therefore, it is not sufficient to observe the effects of weather and roadway features independently as these variables interact with one another.

INVESTIGATION ON RISK PREDICTION OF PEDESTRIAN HEAD INJURY BY REAL-WORLD ACCIDENTS

Head injury is the most common and fatal injury in car-pedestrian accidents. Due to the lack of human test data, real-world accident data is useful for the research on the mechanism and tolerance of head injuries. The objective of the present work is to investigate pedestrian head-brain injuries through real car-pedestrian accidents and evaluate the existed injury criteria. Seven car-to-pedestrian accidents in China were selected from the IVAC (Investigation of Vehicle Accident in Changsha) database. Accident reconstructions using multi-body models were conducted to determine the kinematic parameters associated with the injury and were used to measure head injury criteria. Kinematic parameters were input into a finite element model to run simulations on the head-brain and car interface to determine levels of brain tissue stress, strain, and brain tissue injury criteria. A binary logistic regression model was used to determine the probability of head injury risk associated with AIS3+ injuries (Abbreviated Injury Scale). The results showed that head injury criteria using kinematic parameters can effectively predict injury risk of a pedestrians’ head skull. Regarding brain injuries, physical parameters like coup/countercoup pressure are more effective predictors. The results of this study can be used as the background knowledge for pedestrian friendly car design.

Accident factor analysis based on different age groups via AdaBoost algorithm

A person’s vision, perception, judgment, and operation of a vehicle decline with age. To analyze the influence of age on traffic accidents, we apply the adaptive boosting algorithm (AdaBoost) to investigate the most significant factors for two age groups (older and young driver groups) based on real-world accident data in California. Accident factors include gender, road type, pavement condition, weather, time of day, vehicle behavior, etc., as well as their corresponding subfactors. We first train some weak learners to find importance and then linearly combine those weak learners into a unified stronger learner. The proposed method has several advantages: (1) ability to handle unbalanced data, (2) no requirement on the assumption of data distribution, and (3) being robust for different datasets. Results show that the major factors regarding road safety for older drivers are weather and time of day, while for young drivers are traffic violations and vehicle behaviors.

A STUDY ON MOTORCYCLISTS HEAD INJURIES IN CAR–MOTORCYCLE ACCIDENTS BASED ON REAL-WORLD DATA AND ACCIDENT RECONSTRUCTION

Accident data had shown that as one of the most vulnerable road users, the risk of motorcyclist casualties due to head injuries is extremely high in motorcycle accidents. The objective of this study was to analyze motorcyclist head injury based on epidemiological statistical analysis and car–motorcycle accident reconstruction using real world accident data, and obtain a comprehensive understanding about safety effects of helmets on motorcyclists head injury. In epidemiological data analysis of this study, odds ratios (ORs) was applied to identify differences in injuries between helmeted and unhelmeted motorcyclists. Two vehicle–motorcycle collisions with detailed information were reconstructed by using PC-Crash simulation and MADYMO reconstruction. Furthermore, the head injury reconstruction using MADYMO outputs as boundary conditions was accomplished and analyzed with respect to 1st principal strain, Von Mises stress, coup pressure, countercoup pressure at the cerebrum. The results indicated that unhelmeted motorcyclists were more likely to suffer head injury, serious and fatal injury, and tend to take equal responsibility or more. 1st principal strain was reduced from 0.44 to 0.25 for Case 1, and from 0.16 to 0.10 for case 2 when including a helmet. Von Mises stress were reduced from 30.37[Formula: see text]kpa to 19.51[Formula: see text]kpa for Case 1 and from 3.42[Formula: see text]kpa to 3.03[Formula: see text]kpa for case when including a helmet, which meant a reduction of the risk of concussion. But the motorcyclist in Case 2 experienced a lower percentage increase in coup pressure and countercoup pressure. This study provided comprehensive knowledge on motorcyclists head injuries and the effectiveness of helmets as well as contributing to develop the injury prevention measures and protection devices of motorcyclists.

Effects of passenger car front profile and human factors on pedestrian lower extremity injury risk using German in-depth accident data

ABSTRACTThis study aimed to identify the correlations of pedestrian with moderate to untreatable (AIS2+) lower extremity injury risk with the pedestrian kinematics and vehicle front profile variables using real-world accident data. A subsample of 404 pedestrian accident cases with lower extremity injuries was selected from German in-Depth Accident Study (GIDAS) database to conduct statistical analysis. Variance test and logistic regression analysis was performed to analyze associations between selected variables and risk of pedestrian with AIS2+ lower extremtiy injuries. The results revealed that the predicted variables of the impact speed, pedestrian age and weight, the Lower bumper height (LBH) and the Bonnet leading edge height (BLEH) were statistically significant for AIS2+ lower extremity injuries. The higher impact speed, the higher LBH and BLEH corresponded to a greater likelihood of suffering an AIS 2+ lower extremity injury. The results could provide background knowledge for improving pedestrain protective strategies in passenger vehicle design.

Pedestrian lower extremity injury risk in car-pedestrian collisions

Here, the pedestrian lower extremity injury risk was studied using the real world accident data. For this purpose, 354 cases with pedestrian lower extremity injuries selected from the german in-depth accident study(GIDAS) database were used to conduct the one-way variance analysis to determine the effects of impact speed, pedestrian age, height and weight on the pedestrian serious lower extremity injuries risk. Then, the pedestrian serious lower extremity injury risk model was established. The results showed that the impact speed and pedestrian age are significant factors affecting pedestrian lower extremity serious injuries, but the pedestrian weight and height are not; the risk of injury is positively related to impact speed and pedestrian age; the pedestrian serious lower extremity injury risk reaches 50% when the impact speed is 43 km/h.

Investigation on the injuries of drivers and copilots in rear-end crashes between trucks based on real world accident data in China

This study aimed to investigate the distributing factors affecting the injury severity of drivers and copilots in rear-end crashes between trucks on expressways, based on real world accident data in China. For this purpose, a total of 92 crashes on expressways in Changsha and Zhuzhou areas from 2010 to 2016 were selected. The data was analyzed statistically regarding the corresponding driver, vehicle, roadway and environment characteristics. Firstly, descriptive statistics were conducted to understand the rear-end crash characteristics. Then, ordered logistic regression models were utilized to identify the contributing factors affecting the injury severity of drivers and copilots in rear trucks. Finally, the fatality risk models with respect to relative speed were developed through single logistic regression analyses. The results showed that the probability of driver being fatal decreased if only right fronts of rear trucks were involved in crashes. The probability of copilot being fatal increased if the drivers turned left before crash occurrences or a light truck rear-ended a heavy truck. The relative speed had the highest statistical significance for the truck impact severity. And the copilots were more vulnerable in truck rear-end crashes compared with drivers, with the relative speeds corresponding to 50% probability of being fatal were 32.49 and 35.18 km/h, respectively. The findings from this study provided the theory and data bases for the development of active and passive protection systems.

A Study of Fatality Risk and Head Dynamic Response of Cyclist and Pedestrian Based on Passenger Car Accident Data Analysis and Simulations

Objective: The current study aims to compare the fatality risk of pedestrians and cyclists in urban traffic through an analysis of real-world accident data in China.Methods: First, 438 cases, including 371 pedestrian cases and 67 cyclist cases, were selected as a sample from the accidents collected through an in-depth investigation of vehicle accidents in China. A statistical measurement of the fatality risk with respect to impact velocity was carried out using a logistic regression analysis. Furthermore, 21 pedestrian and 24 cyclist accidents were selected for reconstruction with the MADYMO program. A comparative analysis was conducted based on the results from accident analysis and simulations for the fatality risk and head dynamic response of pedestrians and cyclists.Results: The results indicate that the vehicle impact velocity has a significant relationship with the fatality risk of both pedestrians and cyclists. The fatality risks at 50 km/h are more than twice as high as the risk at 40 km/h and about 5 times as high as that at 30 km/h for both pedestrians and cyclists. Moreover, cyclists suffered slightly lower fatality risk compared to pedestrians. The corresponding vehicle impact velocity is 65.4 km/h for pedestrian with a fatality risk of 50 percent, whereas for cyclists it is 67.6 km/h. In addition, the head impact conditions between pedestrians and cyclists are different.Conclusions: These findings offer potential contributions for establishing a more reasonable speed limit for urban traffic in China and generating strategies for cyclists’ and pedestrians’ head protection.

Benefits of Slope Flattening

The benefits of slope flattening were investigated by simulating accident costs with updated foreslope severities based on real-world accident data collected over a 7-year period in the State of Ohio. Functional classes considered were freeways, rural and urban arterials, and rural and urban local highways. Highways were modeled using the Roadside Safety Analysis Program (RSAP). Highway parameters considered in RSAP were slope steepness, roadway curvature, percent grade, longitudinal length, fill height, and lateral offset to the slope break point. Simulated accident costs were incorporated into a Microsoft Excel spreadsheet, where future users can specify installation costs, which tend to vary significantly from one location to another for slope flattening applications. Each functional class demonstrated slope flattening trends. On freeways and urban arterial highways, slopes should be no steeper than 1V:3H, and the benefit of flatter slopes was minimal. On rural arterial highways, the slope should be no steeper than 1V:4H, and the benefit of flatter slopes was also minimal. On local highways, the steepest slope should be 1V:3H, but the slope should be made as flat as possible because accident costs continued to decrease as the slope was flattened.

Correlation Between Crash Severity and Embankment Geometry

The severity of a roadside feature is often based on survey responses and has tended to emphasize extreme crash events, thereby overestimating the average severity of a particular feature. In this study, severity was related to embankment geometry by examining real-world accident data over a 7-year period. This was done by correlating the number of severe and fatal accidents to the exposure of particular slope geometries. Slope geometry was described by slope steepness and fill height, and its exposure was described by traffic volume and total unshielded mileage. Severity was adjusted for posted speed limits as well. The Roadside Safety Analysis Program (RSAP) was calibrated such that the distribution of severe injury and fatal accidents accurately reflected real-world data. Using this calibrated version of RSAP, the new severity indexes were studied and equations were created to correlate severity index to functional class, fill height, slope steepness, and posted speed limit. The local highway classification provided the highest severity, and the default severity used in RSAP was increased to accommodate this finding. Freeways, rural arterials, and urban arterials experienced reduced severity indexes relative to default values used in RSAP.

High strength steels, stiffness of vehicle front-end structure, and risk of injury to rear seat occupants

Previous research has shown that rear seat occupant protection has decreased over model years, and front-end stiffness is a possible factor causing this trend. In this research, the effects of a change in stiffness on protection of rear seat occupants in frontal crashes were investigated. The stiffness was adjusted by using higher strength steels (DP and TRIP), or thicker metal sheets. Finite element simulations were performed, using an LS Dyna vehicle model coupled with a MADYMO dummy. Simulation results showed that an increase in stiffness, to the extent it happened in recent model years, can increase the risk of AIS3+ head injuries from 4.8% in the original model (with a stiffness of 1000N/mm) to 24.2% in a modified model (with a stiffness of 2356N/mm). The simulations also showed an increased risk of chest injury from 9.1% in the original model to 11.8% in the modified model. Distribution of injuries from real world accident data confirms the findings of the simulations.

A Study of Adult Pedestrian Head Impact Conditions and Injury Risks in Passenger Car Collisions Based on Real-World Accident Data

Objective: The aim of the current study was to study the kinematics of adult pedestrians and assess head injury risks based on real-world accidents. Methods: A total of 43 passenger car versus pedestrian accidents, in which the pedestrian's head impacted the windscreen, were selected from accident databases for simulation study. According to real-world accident investigation, accident reconstructions were conducted using multibody system (MBS) pedestrian and car models under MADYMO environment (Strasbourg University) to calculate head impact conditions in terms of head impact velocity, head position, and head orientation. Pedestrian head impact conditions from MADYMO simulation results were then used to set the initial conditions in a simulation of a head striking a windscreen using finite element (FE) approach. Results: The results showed strong correlations between vehicle impact velocity and head contact time, throw distance, and head impact velocity using a quadratic regression model. In the selected samples, the results indicated that Abbreviated Injury Scale (AIS) 2+ and AIS 3+ severe head injuries with probability of 50 percent were caused by head impact velocity at about 33 and 49 km/h, respectively. Further, the predicted head linear acceleration (head injury criterion, HIC) value, resultant angular velocity, and resultant angular acceleration for 50 percent probability of AIS 2+ and AIS 3+ head injury risk were 116 g, 825, 40 rad/s, 11,368 rad/s2 and 162 g, 1442, 55 rad/s, 18,775 rad/s2, respectively, and the predicted value of 50 percent probability of skull fracture was 135 g. Conclusions: The present study provides new insight into pedestrian head impact conditions in terms of velocity, angle, and impact location based on a number of real-world cases. Therefore, it may perform a critical analysis for current pedestrian head standard tests.

Development of head injury risk functions based on real-world accident reconstruction

The objective of this study is to assess head injury risks of adult pedestrians using computer reconstructions of the accidents. For this purpose, accident reconstructions were developed based on the real-world pedestrian accidents. The impact conditions, which were obtained from the previous study [Y. Peng, C. Deck, J.K. Yang, D. Otte, and R. Willinger, A study of adult pedestrian head impact conditions and injury risks in passenger car collisions based on real world accident data, Traffic Inj. Prev., doi: 10.1080/15389588.2012.733841] including head impact velocity, head position and head orientation, were defined as initial loading conditions in a simulation of the head striking to a windscreen by using a finite element (FE) Hybrid III head model. Logistic regression models were developed to study brain injury risk with respect to injury related variables: the head linear acceleration, HIC value, skull fracture correlate (SFC), resultant angular acceleration and head impact power (HIP). The results from Hybrid III head impact simulations indicated that the predicted head linear acceleration, HIC value, resultant angular acceleration and HIP for 50% probability of AIS 2+ and AIS 3+head injury risk are 116 g, 825, 11368 rad/s2, 37 kW and 162 g, 1442, 18775 rad/s2, 68 kW, respectively, and the predicted values of 50% probability of skull fracture is 135 g. The knowledge from this study could be a prerequisite for developing guidelines to improve pedestrian safety.