Browse Topic: Event data recorders

Items (309)
Find
Development of a Digital Twin of an Electrical Powertrain in a Testbench Environment for Efficiency Estimations2026-01-07447/1/2026
The aim of this work is to develop a modular, real-time-capable digital twin of an electric powertrain based on machine learning (ML)-based model structures and a systematic, component-oriented architecture with a focus on efficiency estimation in test bench environments. The further goal here is to enable virtual testing, which can be used for frontloading and thus both prevent errors and increase the speed of product development. Based on a comprehensive set of measured and derived test bench data, a multi-stage procedure is implemented that integrates data acquisition, physically informed feature selection, modeling at the component and subsystem level, and hybrid coupling strategies. The digital twin captures inverter, electric machine, and mechanical transmission stages and generates consistent predictions of key variables such as torque, speed, power factors, and subsystem as well as overall drivetrain efficiency. The methodology enables a systematic comparison of black box, dark grey box, grey box, and bright grey box architectures with respect to prediction accuracy, information content, and real-time capability. The methodology provided uses new model structures that explicitly integrate physical dependencies while also using ML models to map nonlinear effects. The hybrid architectures presented have been shown to significantly reduce the measurement effort while achieving nearly identical model quality and surpassing purely physics-based models in terms of accuracy, robustness, and real-time capability. For the final bright grey-box architecture, average relative efficiency errors below 1 % are achieved while maintaining real-time execution rates. The study shows that bright grey box-models in particular offer a best-case compromise between the requirements of information content, error quality, and synchronization rate, thus representing a methodological advance over conventional digital twins, which are often created at the component level. The shown methodology provides an implementable framework for digital twins of electric powertrains in industrial test environments.
Kopp, Lennart, Proksch, Daniel, Ockert, Niels, Karthaus, Carsten, Kley, Markus
Improving Regime Recognition Model Validation with Labeled Flight Test Data and Visualization ToolsF-0082-2026-02775/5/2026
Accurate monitoring of helicopter operational usage relies heavily on robust regime recognition algorithms. How-ever, evaluating these approaches is challenging when they operate as opaque, "black boxes", as in the case of machine learning-based models. This paper introduces a comprehensive evaluation framework designed to assess regime recog-nition models from a number of perspectives and investigate anomalies in the predicted regimes. Centered around a high-fidelity data set derived from scripted flight tests covering a complete usage spectrum, the developed method-ology provides a comparative baseline. The analytical suite includes 3D spatial visualization tools for flight path mapping, sequential anomaly detection, and confusion matrix metrics. While applying the labeled data set to other platforms presents inherent limitations in terms of mapping features and regimes appropriately, the integrated toolset successfully exposes weaknesses in the model and highlights gaps in training data. Ultimately, this evaluation frame-work enhances the interpretability of model outputs and builds confidence in the use of regime recognition algorithms.
Cheung, Catherine, Fenev, Nikita, Boldis, Alexander
Advanced Tool for Traffic Crash Analysis: An AI-Driven Multi-Agent Approach to Pre-Crash Reconstruction09-14-01-00335/4/2026
Traffic collision reconstruction traditionally relies on human expertise and, when performed properly, can be incredibly accurate. However, attempting to perform pre-crash reconstruction, i.e., reconstructing the driver and vehicle behaviors that preceded the actual crash, poses significantly more challenges. This study develops a multi-agent artificial intelligence (AI) framework that reconstructs pre-crash scenarios and infers vehicle behaviors from fragmented collision data. We present a two-phase collaborative framework combining reconstruction and reasoning phases. The system processes 277 rear-end lead vehicle deceleration (LVD) collisions from the Crash Investigation Sampling System (CISS; 2017–2022), integrating textual crash reports, structured tabular data, and visual scene diagrams. Phase I generates natural language crash reconstructions from multimodal inputs. Phase II performs in-depth crash reasoning by combining these reconstructions with the temporal event data recorder (EDR). This enables precise identification of striking and struck vehicles while isolating the EDR records most relevant to the collision moment, thereby revealing crucial pre-crash driving behaviors. For validation, we applied it to all LVD cases, focusing on a subset of 39 complicated EDR cases where multiple EDR records per collision introduced possible ambiguity (e.g., due to missing or conflicting data). Ground truth was established via consensus between manual annotations (two independent researchers), with a separate large language model (LLM) used only to flag possible conflicts for re-checking. In the full end-to-end evaluation, the framework achieved 100% accuracy across all 4155 trials (277 cases × 5 runs × 3 models), with three reasoning models producing identical outputs, confirming that performance derives from the structured prompt design rather than model-specific characteristics. In contrast, research analysts without specialized reconstruction training achieved 92.31% accuracy on the same 39 complex cases. In separate ablation experiments on the 39 complicated EDR cases, where one randomly selected Phase I output from the full end-to-end evaluation was fixed as the unified input for Phase II and each model was tested with 10 independent runs, removing the structured reasoning anchors reduced case-level accuracy from 99.7% to 96.5%, with errors spreading from a single output type to multiple analytical dimensions. The system maintained robust performance even when processing incomplete data. This zero-shot evaluation, conducted without any domain-specific training or fine-tuning, demonstrates that the framework’s effectiveness stems from its multi-agent architecture and prompt engineering, offering a scalable approach for AI-assisted pre-crash analysis.
Xu, Gerui, Chen, Boyou, Guo, Huizhong, LeBlanc, Dave, Kusari, Arpan, Yarbasi, Efe, Ahmed, Ananna, Sun, Zhaonan, Bao, Shan
Visual–Language Model–Driven Annotation and Analysis of Maneuver Data for Enhanced Vehicle Safety09-14-01-00325/4/2026
Vehicle maneuver data are essential for perception and planning in advanced driver-assistance systems (ADAS) and automated driving systems (ADS). While high-quality annotations improve machine-learning performance, existing maneuver datasets remain fragmented, labor-intensive to annotate, and inconsistent in semantic richness. Challenges persist in scalability, interpretability, and contextual labeling. This article establishes a structured framework for maneuver data analysis by combining a systematic review of existing resources with the development of a new multimodal dataset. First, we conduct a systematic review of publicly available datasets such as HDD, KITTI, BDD-X, D2CAV, Brain4Cars, DrivingDojo, and the Driving Behavior Database. We further evaluate the data modality and sensor configurations including event data recorders, onboard logging systems, and smartphone sensing. We then propose the Matt3r Data Collection System with modern metadata management, which integrates video, GPS, and IMU signals into temporally coherent clips. Next, we outline the limitations of traditional annotation approaches, which rely on manual labeling and rule-based methods. To address the limitations of traditional manual and semi-automated labeling, we propose a Vision–Language Model (VLM)–driven annotation pipeline. VLMs generate maneuver categories and causal explanations through prompt-based reasoning, with selected outputs refined through human-in-the-loop verification. Finally, we propose an annotation quality evaluation based on accuracy, inter-annotator agreement, credibility, consistency, and efficiency gain. In summary, this article bridges the gap between the environment perception requirements of existing ADAS and ADS systems and the developing capabilities of generative artificial intelligence. By providing a novel and scalable research approach for AI-driven maneuver data annotation and analysis, this article supports data engineering efforts for both research and practical applications aimed at enhancing vehicle safety.
Bai, Ling, Yuan, Chongyu, Osman, Islam, Lin, Zirui, Mirab, Ghazal, Saheb, Amir, Parnian, Neda, Shapiro, Evgeny, Shehata, Mohamed S., Liu, Zheng
Regional-Specific Patterns of Abdominal Injury in Belted Male and Female Frontal Crash Occupants: Insights from Redefined Anatomical Mapping09-14-01-00025/4/2026
Research Question/Methods The study examined abdominal injuries of 87 belted occupants in CIREN frontal crashes for sex-based differences in abdominal injury patterns. It introduced a more anatomically detailed method for identifying injury locations in an abdominal-pelvic region that includes skeletal structures. The study introduces and applies a novel Abdominal New Injury Severity Score (AbNISS) to address limitations of traditional AIS coding in capturing sex-based differences in injury patterns. The operative reports/EDR/imaging data in CIREN cases enabled identification of sex-specific crash outcomes. The dominant analytical motif is Bertrand Russell’s knowledge by acquaintance and definite descriptions. Results Females had a higher rate of moderate to severe abdominal injuries than males: Only females sustained AIS 5 injuries, lumbar Chance fractures, posterior pelvic arch injuries, and more AIS 2, 3, and 4 injuries, with more injuries in superior-mid, left-superior, and medial-mid-abdominal zones. Males had more in the medial-inferior zone. 13 females had muscle ruptures. Four had combined muscle and Morel–Lavallée injuries, and four had Morel–Lavallée injuries alone. 13 of 14 males had muscle ruptures only. Pelvic morphology: Statistically significant (p < 0.05) sex-based difference in pelvic aspect ratios: Females: 1.37 ± 0.053, with the male ratio of 1.28 ± 0.079. By incorporating anatomical precision and enhanced injury mapping, it supports the development of more representative ATD/human body models. Discussion and Limitations Limitations of the study include its retrospective design, possible inconsistencies in clinical documentation, and challenges in applying AbNISS coding to non-CIREN datasets due to specificity constraints.
Halloway, Dale, Curry, William, Somasundaram, Karthik, Pintar, Frank
The Effects of Collision-Related Power Loss on Toyota Safety Sense 3.0 Event Data2026-01-05484/7/2026
Toyota vehicles equipped with Toyota Safety Sense (TSS) can record detailed information surrounding various driving events. Often, this data is employed in accident reconstruction to better understand the dynamics of a collision. TSS data is comprised of three main categories: Vehicle Control History (VCH), Freeze Frame Data (FFD), and image records. During an event, it is possible that a vehicle undergoes a catastrophic power loss from the damage sustained during the event. In this paper, the effects of sudden power loss on the VCH, FFD, and images are studied. Events are triggered on a TSS 3.0 equipped vehicle by driving toward a stationary target. After system activation, a total power loss is induced, triggered on the instrument cluster “BRAKE” alert message, at various delays after activation. This testing studies various signals recorded across VCH, FFD and image data including vehicle speed and time and date. Results show that there is a minimum time to record after system activation to record any related data. Power losses which occur before this minimum time record no data and losses after the minimum time record incomplete data sets. Studied vehicle signals were all in general agreement with measured values.
Getz, Charles, Yeakley, Adam, DiSogra, Matthew
Pre-Crash Scenario Analysis and Generation for Integrated Safety System Virtual Testing with Real Accident Data2026-01-00624/7/2026
Integrated active and passive safety protection systems have made substantial contributions to reducing traffic accidents and mitigating human injuries. However, assessing such systems through vehicle collision tests is limited, as this approach cannot cover the wide range of accident scenarios. To address this gap, identifying and generating representative pre-crash scenarios from real-world accidents provides key boundary conditions for the setup of virtual test scenarios. In this study, we used the Future Mobile Traffic Accident Scenario Study (FASS) dataset to reconstruct 112 two-wheeler accidents. For each case, we extracted pre-crash dynamic information, static attributes, and environmental context. An autoencoder was employed to encode high-dimensional features of scenarios, and K-means clustering was applied to categorize the accidents into eight representative pre-crash scenarios. For each scenario, we examined the motion states of participants and further compared the behaviors and injuries between motorcycle and bicycle accidents. The results show that motorcycles have a higher average pre-crash speed than bicycles (22.4 vs 16.2 km/h), resulting in a greater proportion of Maximum Abbreviated Injury Scale (MAIS) 4+ cases (35.6% vs 18.9%). Furthermore, a Gaussian Mixture Model was applied to fit the scenario features. This model was then used to randomly generate the initial conditions of pre-crash scenarios, including the positions, speeds, and yaw angles of vehicles and two-wheelers. The proposed scenario generation method was first applied to pre-crash scenario construction for integrated safety systems tests, and it can create diverse, statistically grounded scenarios that reflect real-world accident distributions. These scenarios can broaden the test scope of integrated safety systems and support their implementation.
Wang, Guojie, Gao, Xin, Liu, Siyuan, Liu, Jiaxin, Li, Quan, Shi, Liangliang, Nie, Bingbing
Inertial Pedal Displacement and Brake Switch Activation in Frontal Crash Events2026-01-05414/7/2026
This paper presents research into the inertial displacement of brake pedals and the subsequent activation of brake light switches during crash events. In certain scenarios, such as multiple-impact crashes or crashes with pre-impact interactions such as curb strikes or sideswipes, inertial forces alone may generate sufficient brake pedal movement to trigger the brake switch, activating the brake lights. Such signals may be recorded by an Event Data Recorder (EDR) or observed by witnesses and incorrectly interpreted as an indication of intentional driver braking. To investigate this phenomenon, HYGE sled tests were performed using brake pedal assemblies and associated components from a Toyota Tacoma pickup truck and a Cadillac DeVille passenger sedan. The assemblies were subjected to acceleration pulses simulating a frontal impact, with high-speed video used to capture brake pedal displacement and brake light activation. The tests demonstrated that inertial loading from a pulse with a delta-V (change in velocity) as low as 12 mph could result in momentary brake light activation due to pedal displacement from inertial forces. Increasing the magnitude of the acceleration pulse produced greater displacement of the brake pedal and extended the duration of the brake light activation. An example is presented that demonstrates inertial pedal movement in a full-scale vehicle test conducted by striking a curb, which resulted in brake light activation with a delta-V considerably less than 12 mph. Additionally, a field survey of 50 passenger vehicles found that the brake switch activation threshold ranged from 0.25 to 0.56 inches of pedal travel for 80% of the vehicles measured. These findings indicate that relatively small crash accelerations and durations can produce sufficient inertial pedal movement to activate brake lights and that only minimal pedal displacement is required in most vehicles.
Walker, James, Duran, Amanda, Barnes, Daniel, Osterhout, Aaron, Clayton, Aidan
Evaluation of PAEB Performance of a GM Vehicle Using EDR and Ground Truth Data2026-01-00544/7/2026
In order to determine the on-board EDR data recording characteristics of a GM vehicle, a 2023 GMC Sierra Denali was tested in several Pedestrian Automatic Emergency Braking (P-AEB) scenarios. Using a variety of test tools, including the STRIDE robotic platform and its onboard data systems, a GPS/IMU installed in the vehicle, and several camera units, the vehicle was put into collision imminent scenarios in which the crash avoidance systems were actuated. The flags in the EDR data, the order in which EDR events were written, and the correlation between the EDR and data recorded by the aforementioned external acquisition systems were examined for each test case. Testing was done in both forward and reverse scenarios and at low speeds only. These results provide a picture of the current state of the additional data available in current EDRs installed on GM vehicles equipped with P-AEB capability, as well as an insight into the accuracy and meaning of that data which should prove beneficial to the accident reconstruction community, among others.
Bartholomew, Meredith, Arnett, Michael, Guenther, Dennis
Validation of a Legacy-Compatible Vehicle Communication Interface for the Retrieval of Crash Data from Passenger Vehicles2026-01-50092/3/2026
The discontinuation of the Bosch CANplus vehicle communication interface (VCI), in October of 2023, created a significant gap in crash data retrieval workflows for many legacy vehicles that have not been transitioned to the newer CDR 900 VCI. A substantial portion of CDR-supported vehicles continue to rely on legacy CANplus-based communications, including the majority of platforms from the mid-1990s through the early 2010s as well as select late-model vehicles from manufacturers such as Chevrolet that still require CANplus-compatible communication for imaging of crash data. This study evaluates a legacy-compatible VCI designed to operate within existing Bosch CDR software, adapter, and cable infrastructures. Thirty controlled test scenarios were performed across ten vehicle brands using airbag control modules (ACMs), a powertrain control module (PCM), and selected modules that had undergone prior chip-swap procedures. Test conditions included DLC imaging, DTM imaging, and back-powering through the vehicle fuse box. Each module was imaged using the discontinued Bosch CANplus VCI and the compatible Crash Pulse Technologies Crash Data eXtractor (CDX) VCI under identical procedures, and the resulting datasets were compared at the character level. All scenarios produced complete and consistent outputs between devices, with differences limited to expected non-critical data such as timestamps and user-entered comments. The findings demonstrate that the evaluated VCI provides reliable and repeatable crash data retrieval performance equivalent to the discontinued CANplus across the legacy workflows tested. The validation framework aligns with accepted reliability principles referenced in Daubert considerations and supports the use of the CDX VCI in forensic crash investigations and related technical applications.
Zeitler, Jason Paul
A Large-Scale Characterization of PyCrash Collision Model Behavior Using Real-World Collision Data2026-01-50041/15/2026
PyCrash is an open-source collision simulation software package that includes a formulation of the Carpenter–Welcher collision model. Upon its release, PyCrash was accompanied by a companion paper that described its functions and provided preliminary validation results against staged collisions. However, the collisions investigated in the original report were limited to a single type of alignment. The purpose of this study was to characterize PyCrash collision model behavior against EDR data collected from a heterogeneous cohort of real-world two-vehicle collisions. PyCrash simulations were informed by the published vehicle geometries, crush profiles, and available pre-impact EDR data; simulation outputs were compared to EDR data, which served as the surrogate for “ground truth” with respect to the collision mechanics. Simulation settings were tuned to the specifics of each crash, based on previous published work and engineering judgement. Using optimized inputs for each crash, PyCrash simulations produced collision findings that were, on average, within 1 mph of the recorded EDR data. Average disagreement, normalized to EDR data, was 10% or less. The use of a large cohort of published real-world crash data indicates that PyCrash simulations can be tuned to model collision mechanics within the known error rate of EDR data for a range of crash alignments.
Fischer, Patrick, Cormier, Joseph, Watson, Richard
An Interpretable Deep Learning Decision-Making Framework for Autonomous Lane-Change Inspired by Human Driving Experiences2025-01-733912/31/2025
Lane change plays a critical role in autonomous driving and directly affects traffic safety and efficiency. Although deep learning-based lane-change decision-making frameworks have achieved promising results, they still face fundamental challenges in producing human-consistent and trustworthy behavior, mainly due to: 1) Inadequate psychology-informed personalization, as most frameworks focus on physical variables but neglect psychological factors (e.g., risk tolerance, urgency), limiting their ability to capture individual differences in lane-change motivations. 2) Limited holistic understanding of traffic context, most frameworks lack consideration of high-level and interpretable indicators (e.g., traffic pressure) in comprehensively assessing dynamic traffic scenarios, limiting their capacity for human-like contextual understanding. 3) Lack of transparent and interpretable decision logic, as many frameworks operate as black boxes with opaque reasoning processes, hindering human-aligned explanation, weakening user trust, reducing accident traceability, and impeding model refinement. To this end, a policy-oriented contextual-reasoning fuzzy neural network (POCR-FNN) is proposed as a deep learning-based decision-making framework for personalized and interpretable autonomous lane-change. First, we develop a psychology-informed driving style classification by learning distinct fuzzy membership functions to enable style-specific policy learning. Second, we design a human-inspired local interaction-aware module that estimates traffic tension by combining interaction salience and contextual risk, enhancing contextual understanding. Finally, we integrate fuzzy logic with a deep learning-based policy network to enable rule-level decision reasoning with real-time interpretability and transparent traceability. Extensive experiments on multiple public highway and urban datasets demonstrate that POCR-FNN achieves state-of-the-art performance while significantly improving personalization and interpretability across various driving styles and scenarios.
Chen, Yanbo, Chen, Jiaqi, Yu, Huilong, Xi, Junqiang
Pedestrian Accident ReconstructionR-59010/30/2025
As technology advances, so too must the methods we use to investigate and understand pedestrian collisions. Pedestrian Accident Reconstruction provides a much-needed update to the field—one that reflects the dramatic rise in digital evidence and the growing role of simulation in modern reconstructions. While earlier texts laid the foundation for understanding pedestrian impacts, they were written before the widespread use of video footage, onboard vehicle sensors, and powerful simulation software. This book addresses that shift. It introduces updated methods that incorporate event data recorders, surveillance video, dash cams, and advanced simulation tools like the PC-Crash multibody pedestrian model. These technologies allow for more precise, repeatable, and illustrative reconstructions than ever before. The book highlights when and how simulation can enhance an investigation—especially in complex scenarios such as roof vaults, sideswipes, or collisions that defy conventional equations. Though PC-Crash is the primary tool used throughout, the insights and principles are broadly applicable across simulation platforms. Written by experienced practitioners, this is not just a guide to software—it’s a thoughtful, practical resource for applying modern tools to real-world cases. Whether you're new to the field or looking to refine your practice, this book offers an essential perspective for today’s reconstruction professional.
Rose, Nathan, Carter, Neal
Space Laser Tech Maps Mysterious Nearshore TerrainTBMG-535468/1/2025
Despite all the technological evolution in navigation, waters just off coastal shores around the globe have remained a black box. That is, until researchers from The University of Texas at Austin and Oregon State University developed a new technology that uses satellites in space to map out these tricky areas.
Pulse and Polynomial Functions to Predict Vehicle Acceleration Using Event Data Recorder Delta-V in Frontal and Rear-End Collisions2025-01-50466/27/2025
The lack of recorded acceleration and limited Delta-V (ΔV) resolution in many vehicle event data recorders necessitates the development of a method to predict continuous vehicle acceleration based on ΔV responses. This study developed a method of obtaining continuous acceleration by regressing pulse functions (triangular, half-sine, haversine) and polynomial functions (orders 3–6) to a ΔV curve and deriving the corresponding acceleration–time curve. The effectiveness of this method was demonstrated using real-world ΔV response data from front and rear-end collisions. Comparisons were performed between peak and average acceleration values from each front and rear-end crash pulse. Results indicated that a triangular pulse function predicted similar peak acceleration values to the vehicle’s actual acceleration in frontal and rear-end impacts. Average acceleration in frontal impacts was best predicted utilizing a fifth-order polynomial, while a sixth-order polynomial demonstrated the best predictive ability for rear-end impacts. Obtaining equations for vehicle ΔV and acceleration is crucial in assessing impact severity due to the vehicle’s dynamic response.
Westrom, Clyde, Adanty, Kevin, Shimada, Sean D.
Analysis of the Event Data Recorder Vehicle System Data of a GM Vehicle Intelligence Platform Airbag Control Module (SDM50)2025-01-87144/1/2025
The primary function of an Airbag Control Module (ACM), referred to as the Sensing and Diagnostic Module (SDM) by General Motors (GM), is to detect crashes, discriminate crashes, evaluate crash severities, deploy the appropriate restraints, including airbags and pretensioners, and perform system diagnostics. A secondary function of the SDM is to act as an Event Data Recorder (EDR) which records data during the time periods just prior to (pre-crash) and during a crash event. This data consists of restraint and vehicle system data which is collected, processed, and stored in the EDR. Data stored in the EDR is intended to be retrieved after a crash. This data provides operational information on the vehicle’s occupant protection system and other vehicle systems to assess system performance, aid in crash reconstruction, and support improved vehicle safety. A series of vehicle test maneuvers were conducted while injecting a non-deployment crash pulse directly into the SDM to cause the SDM to record an event with related restraint and vehicle system (pre-crash) data. These tests include a variety of constant speed, acceleration, braking, and steering maneuvers. During these maneuvers, vehicle system data was recorded directly from the serial data bus using a passive monitoring tool, tapped sensors, video cameras, and independent onboard instrumentation and that data was compared with the data recorded within the SDM EDR. This paper addresses the operation and accuracy of the vehicle system data recorded for a crash event by the SDM, specifically the SDM50 and its utilization of the GM Vehicle Intelligence Platform (VIP) vehicle serial communication bus. Evaluation of this data provides an understanding of the accuracy of the vehicle system data recorded for a crash event by the SDM.
Smyth, Brian, Crosby, Charles L, Bickhaus, Ryan, Smith, James, Edmunds, Dustin, Floyd, Donald, Modi, Vipul, Outlaw, RaShawndra D., Wright, Jeff
A Model for the Design of Automated Vehicle Event Data Recorders2025-01-86784/1/2025
When vehicle accidents occur, investigators rely on event data recorders for accident investigations. However current event data recorders do not support accident investigation involving automated or self-driving vehicles when there is state information that needs to be recorded, for example ADS modes, changes in the ODD that the vehicle operates under, and the various states of vehicle features such as intelligent cruise control, automated lane changes, autonomous emergency braking, and others. In this paper, we propose a model to design new types of event data recorders that supports accident investigations involving automated vehicles when there is state information to be recorded. The model is generic enough to be adapted to any automation level and any set of automated vehicle functional features. The model has been instantiated to a specific ADAS system.
Pimentel, Juan
Examination of the Data Structures of the Bendix® Data Recorder2025-01-87124/1/2025
Bendix® EC-80™ and certain EC-60™ ABS control units contain an event data recorder called the Bendix® Data Recorder (BDR). Raw BDR data is obtained using commercially available software, however, the translation of the raw data into an event report has only been performed by the manufacturer. In this paper, the raw data structures of the commercially available datasets are examined. It is demonstrated that the data follows uniform and repeatable patterns. The raw BDR data is converted into a conventional report and then validated against translation reports performed by the manufacturer. The techniques outlined in this research allow investigators to access and analyze BDR records independently of the manufacturer and in a way previously not possible.
DiSogra, Matthew, Hirsch, Jeffrey, Yeakley, Adam
Neural Network Modeling of Black Box Controls for Internal Combustion Engine Calibration2024-01-29957/2/2024
The calibration of Engine Control Units (ECUs) for road vehicles is challenged by stringent legal and environmental regulations, coupled with short development cycles. The growing number of vehicle variants, although sharing similar engines and control algorithms, requires different calibrations. Additionally, modern engines feature increasingly number of adjustment variables, along with complex parallel and nested conditions within the software, demanding a significant amount of measurement data during development. The current state-of-the-art (White Box) model-based ECU calibration proves effective but involves considerable effort for model construction and validation. This is often hindered by limited function documentation, available measurements, and hardware representation capabilities. This article introduces a model-based calibration approach using Neural Networks (Black Box) for two distinct ECU functional structures with minimal software documentation. The ECU is operated on a Hardware-in-the-Loop (HiL) rig for measurement data generation. To build surrogate models of these ECU functions, Neural Network model inputs are allocated categorized into two categories: function inputs as perceived by the logic level (White Box) software function, and curve/map fitting features representing the adjustment variables of the ECU function. Factors influencing surrogate model accuracy such as, Neural Network hyperparameter optimization, input space amount and distribution as well as the parameter adjustment is investigated. Results show an increase in accuracy with the increasing number of implemented parameters, as well as the scalability of ECU function model representation with measurement data. In addition to calibration purposes, the presented function representation method facilitates the use of plant models to replace time-consuming function construction and validation.
Meli, Matteo, Wang, Zezhou, Bailly, Peter, Pischinger, Stefan
Analysis of the Event Data Recorder (EDR) Function of a GM Active Safety Control Module (EOCM3 LC)2024-01-28884/9/2024
The Advanced Driver Assistance System (ADAS) is a comprehensive feature set designed to aid a driver in avoiding or reducing the severity of collisions while operating the vehicle within specified conditions. In General Motors (GM) vehicles, the primary controller for the ADAS is the Active Safety Control Module (ASCM). In the 2013 model year, GM introduced an ASCM utilizing the GM internal nomenclature of External Object Calculation Module (EOCM) in some of their vehicles produced for the North American market. Similar to the Sensing and Diagnostic Module (SDM) utilized in the restraints system, the EOCM3 LC contains an Event Data Recorder (EDR) function to capture and record information surrounding certain ADAS or Supplemental Inflatable Restraint (SIR) events. The ASCM EDR contains information from external object sensors, various chassis and powertrain control modules, and internally calculated data. This event data includes date and time, GPS location, driver inputs and vehicle responses, and information regarding ADAS objects of interest. This paper addresses the operation and accuracy of the EDR data recorded by an ASCM, specifically the GM EOCM3 LC, and its utilization of the GM Vehicle Intelligence Platform (VIP) inter-module serial communication bus. A series of vehicle Automatic Emergency Braking (AEB) test maneuvers were conducted, triggering the ASCM EDR function. The vehicle dynamic state was independently monitored and recorded by onboard instrumentation and compared to the ASCM recorded data. Evaluation of these data provides a better understanding of the accuracy and timing of this event data recorded by the General Motors' ASCM.
Bare, Cleve, Skiera, Jason, Smyth, Brian, Beetham, Tommy, Floyd, Donald, Koo, Winston, Newell, Devin
Accuracy of 2016-2022 EDRs in IIHS Crash Tests2024-01-28904/9/2024
Event data recorders (EDRs) were harvested and imaged after Insurance Institute for Highway Safety (IIHS) 56 km/hr frontal and 64.4 km/hr frontal offset crashes of 15 different brands of 2016-2022 vehicles. The speed and delta-V in the EDR were compared to reference instrumentation. Speed data was accurate within the generally accepted range of +/-4%. The 40% overlap tests had generally similar vehicle kinematics, and their delta-Vx data was accurate. However, there was a much greater variance in the small (25%) overlap tests. Some outliers in the small overlap delta-Vx tests required further analysis using overhead video analysis. The video analysis more closely matched the EDR recorded values. These offset tests create significant post-crash rotation, and both EDR and IIHS instrumentation were affected by their location away from the center of gravity. The Y-axis was affected much more than the X-axis. The data scatter in Y-axis was significant, particularly in the IIHS reference instrumentation. Quantitative corrections were calculated and reduced the data set differences, but did not bring every crash test into agreement.
Ruth, Richard, King, Charles, Rich, Andrew, Sadrnia, Hamed
Study on a Method for Reconstructing Pre-Crash Situations Using Data of an Event Data Recorder and a Dashboard Camera2024-01-28914/9/2024
When investigating traffic accidents, it is important to determine the causes. To do so, it is necessary to reconstruct the accident situation accurately and in detail using objective and diverse information. We propose a method for reconstructing the accident situation (“reconstruction method”) which consists of rebuilding the situation immediately before the collision (“pre-crash situation”) using data collected during that time by an event data recorder (EDR) and a dashboard camera (DBC) onboard one or both of the vehicles involved. First, the vehicle’s traveling trajectory was integrally calculated using the vehicle speed and yaw rate recorded by the EDR, each point along the trajectory being linked to the EDR data. After being combined with the DBC’s video data, the trajectory was projected onto the road surface around the accident site, which allowed us not only to display on a single road map the vehicle’s traveling trajectory, but also to provide, on each point along the trajectory, diverse information from EDR data and freeze-frame pictures of the road ahead from the DBC. The reconstruction method was then applied to a real-world accident involving the collision of two vehicles at an intersection. The analysis of the reconstructed accident situation showed that the Advanced Emergency Braking System (AEBS) of one of the vehicles was not designed to be activated in this collision scenario. Using this reconstruction method, objective and diverse pieces of information (collected by EDRs and DBCs) and traveling trajectories obtained from traffic accident investigations were projected onto a road map in a chronological and integrated way to reconstruct the pre-crash conditions as accurately and in as much detail as possible. Further, if one or both vehicles involved in the accident had the EDR’s pre-crash data and the DBC’s images regarding an advanced driver assistance system (ADAS), this reconstruction method allows us to easily understand the pre-crash situation through the diagram that reconstructed how the ADAS was working (EDR data) and the position of the other vehicle ahead immediately before the accident (DBC images).
Matsumura, Hideki, Sugiyama, Motoki, IWATA, Takekazu
Simulation of Vehicle Speed Sensor Data for Use in Heavy Vehicle Event Data Recorder Testing2024-01-28894/9/2024
Heavy Vehicle Event Data Recorders (HVEDRs) have the ability to capture important data surrounding an event such as a crash or near crash. Efforts by many researchers to analyze the capabilities and performance of these complex systems can be problematic, in part, due to the challenges of obtaining a heavy truck, the necessary space to safely test systems, the inherent unpredictability in testing, and the costs associated with this research. In this paper, a method for simulating vehicle speed sensor (VSS) inputs to HVEDRs to trigger events is introduced and validated. Full-scale instrumented testing is conducted to capture raw VSS signals during steady state and braking conditions. The recorded steady state VSS signals are injected into the HVEDR along with synthesized signals to evaluate the response of the HVEDR. Brake testing VSS signals are similarly captured and injected into the HVEDR to trigger an event record. The results show that HVEDR event records can be precisely and repeatedly triggered using synthesized signals. The synthesized signals produce identical data within the HVEDR compared with the original signals. Finally, various use cases for the method are discussed.
DiSogra, Matthew C., Getz, Charles, Patel, Amit, Grimes, Wesley, Plant, David, Wilcoxson, Greg
Comparing Event Data Recorder Data (EDR) in Front/Rear Collisions from the Crash Investigation Sampling System (CISS) Database2024-01-28924/9/2024
The accuracy of collision severity data recorded by event data recorders (EDRs) has been previously measured primarily using barrier impact data from compliance tests and experimental low-speed impacts. There has been less study of the accuracy of EDR-based collision severity data in real-world, vehicle-to-vehicle collisions. Here we used 189 real-world front-into-rear collisions from the Crash Investigating Sampling System (CISS) database where the EDR from both vehicles recorded a severity to examine the accuracy of the EDR-reported speed changes. We calculated relative error between the EDR-reported speed change of each vehicle and a speed change predicted for that same vehicle using the EDR-reported speed change of the other vehicle and conservation of momentum. We also examined the effect of vehicle-type, mass ratio, and pre-impact braking on the relative error in the speed changes. Overall, we found that the common practice of using the bullet vehicle’s EDR-reported speed change to estimate the target vehicle’s speed change is reliable and can be adjusted for some vehicle types and mass ratios to improve the accuracy of the target vehicle’s estimated speed change. We also found that about 10-13% of EDR-reported speed changes may have errors larger than ±10%.
Fix, Ryan, Wilkinson, Craig, Siegmund, Gunter P.
Event Data Recorder Trigger Probability in the Crash Investigation Sampling System Database2024-01-50272/23/2024
The objective of this study was to quantify the field performance of passenger vehicle event data recorders (EDRs) in recording data into non-volatile memory at the 8 km/h delta-v (Δv) trigger threshold specified by Title 49, Part 563 of the Code of Federal Regulations (Part 563). Part 563 applies to passenger vehicles manufactured on or after September 1, 2012. The trigger threshold is distinct from the threshold required to deploy an airbag. Events meeting the trigger threshold will cause data to be preserved on the EDR even if airbags are not deployed. This is the first study to quantify EDR trigger threshold performance. This data is valuable in the evaluation of sub-airbag deployment crashes. The study was accomplished via analysis of EDR and reconstructed Δv data from 3,960 cases in the Crash Investigation Sampling System (CISS) database maintained by the National Highway Traffic Safety Administration (NHTSA). The binary presence or non-presence of an event on the EDRs of vehicles exposed to a collision was compared to the CISS reconstructed Δv for each vehicle. Logistic regression models were developed to predict the probability of an event present on the EDR at the trigger threshold. We found that vehicles manufactured by Toyota had lower Δv thresholds compared to other manufacturers. For Toyota manufactured vehicles, the probability of an EDR event at 8 km/h ranged from 87% to 99%. EDR event probabilities for non-Toyota vehicles in vehicle-to-vehicle collisions ranged from 58% to 93% for Part 563 compliant vehicles and 33% to 83% for pre-Part 563 vehicles. The persistence of EDR events in memory was analyzed using the number of ignition cycles present between events and imaging of the EDR, finding average duration of 3,595 ignition cycles for pre-existing EDR events unrelated to the CISS case.
Watson, Richard A., Bonugli, Enrique, Greenston, Mathew, Santos, Erick, Martinez, Jonathan
Study of Vehicle-Based Metrics for Assessing the Severity of Side Impacts09-12-01-000410/30/2023
A research program has been launched in Iran to develop an evaluation method for comparing the safety performance of vehicles in real-world collisions with crash test results. The goal of this research program is to flag vehicle models whose safety performance in real-world accidents does not match their crash test results. As part of this research program, a metric is needed to evaluate the severity of side impacts in crash tests and real-world accidents. In this work, several vehicle-based metrics were analyzed and calculated for a dataset of more than 500 side impact tests from the NHTSA crash test database. The correlation between the metric values and the dummy injury criteria was studied to find the most appropriate metric with the strongest correlation coefficient values with the dummy injury criteria. Delta-V and a newly created metric T K 200 Y , which is an indicator of the kinetic energy transferred to occupants in a 200 ms time interval and in the lateral direction, were found to be the most appropriate metric for assessing the crash severity of side impacts with strong correlation coefficients with head injury criteria such as HIC36 and HIC15, resultant spinal acceleration, and moderate correlation coefficients with average rib deflection and abdominal forces. Due to the need to calculate the metric based on EDR measurements, T K 200 Y was chosen as the side impact severity metric for the research program.
Sadeghipour, Emad
Research of Vehicle Behavior Classification of Off-Road Motorcycles Using Gradient Boosting2023-01-181710/24/2023
Machine learning is used for the research and development of ITS services and the rider assistance for on-road motorcycle racing. Meanwhile, rider assistance systems for off-road motorcycles have yet to be developed, partly due to the complexity of the measurement conditions, as described in the previous paper. This research aims to create a reliable AI which is capable of classifying typical jump behaviors in off-road riding by machine learning to create a rider assistance system for off-road motorcycles. Motorcycle manufacturers and certain research institutes use motion sensors to collect data, but the data is obtained from a limited number of vehicles and riders. The creation of a rider assistance system requires a large amount of validation data. Furthermore, it is desirable to achieve the target with data that can be measured in mass-produced vehicles, which will make it possible to collect data even from general users. In addition, recent machine learning models are black boxes because it is difficult for people to understand the entire process, and it is necessary to evaluate the validity of the results. The approaches are as follows. (1) Using data that can be measured in mass-produced vehicles, the number of features was increased as a preprocessing step. (2) The validity of the machine learning model was evaluated by focusing on the SHAP value, one of the XAI techniques. As a result, (1) classification ability has improved. (2) correspondence between the number of features with large SHAP values and physical phenomena has been obtained. In other words, it has been confirmed that appropriate number of features have been selected for classification. These results have indicated that the created AI has a certain level of classification ability and that the judgment results can be trusted.
Uto, Yuki, Tokunaga, Hisato, Inaba, Taichi, Higashi, Takayuki
AI/ML Digital Twins of Hardware-in-the-Loop Systems2024-01-41058/10/2023
Proprietary, black box, and other hard-to-model subsystems are a leading source of schedule and labor cost across simulation supported analysis and lifecycle management. Using AI/ML technologies to rapidly develop and deploy digital twins of Hardware in the Loop (HWIL) and software systems reduces the Non-Recurring Engineering (NRE) in Modeling and Simulation (M&S) and supports validation of existing software digital twins. This approach also allows for portability of obsolete or proprietary components into a broader range of simulations or applications without exposing critical technologies. We present results of multiple case studies applying AI to black box components of interest to the ground vehicle community.
Colley, Wesley N., Banyai, Joel, Gordy, Joshua, Mills, Matthew, Warren, Randall
MBSE Strategies: Avoiding Cyclic Project Usage Dependencies by using Black Box and White Box Models2024-01-40838/10/2023
Every digital engineering framework and modeling approach will include benefits and concerns. It is important to customize the response, within reason and based on the available resources, to the needs of the project and contract. For this case, the consideration of a large, singular model was overturned for a distributed model. The potential for a cyclic usage, which can be catastrophic in both performance issues and data loss, was mitigated by an innovative approach that allowed for two (2) systems models – one (1) Black Box and one (1) White Box – using a novel model federation strategy. The concerns of having two (2) system models were mitigated via acceptance and understanding that each system model would play its part appropriately based on model function, system development, and contract deliverables.
Kolligs, Jason, Masterson, Stuart, Thome, Eric, Knutson, Christian
Frontal Crash Reconstruction Compared to Event Data Recorders in the Crash Investigation Sampling System Database and the Effect on Injury Risk Models2023-01-50437/17/2023
This study compares statistical models for frontal crash injuries based on delta-v data reported by the vehicle event data recorder (EDR) with injury probability models based on delta-v reconstructed by Crash Investigation Sampling System (CISS) investigators. Injury probabilities and their follow-on use in advanced automatic crash notification (AACN) systems have traditionally been based on delta-v obtained through accident reconstruction of field crashes in the National Automotive Sampling System Crash Data System (NASS-CDS) database. Field delta-v from EDRs in the CISS database is an alternative source of information for crash injury probability modeling. In this study, frontal impact injury risk probabilities computed from EDR and reconstructed delta-v were compared. All data came from the years 2017–2021 of the CISS database, which contains EDR downloads and also reconstructed delta-v using crush measurements and NHTSA’s WinSmash software. On average, CISS reconstructions overestimated delta-v below 16 kph and underestimated delta-v above 16 kph when compared to EDR delta-v. CISS also records detailed injury data using the 2015 Abbreviated Injury Scale (AIS) developed by the Association for the Advancement of Automotive Medicine (AAAM). Statistical analysis was performed using logistic regression to calculate MAIS 1+, MAIS 2+, MAIS 3+, and ISS 16+ injury probabilities. EDR-derived injury probabilities showed a lower risk for serious injury at delta-v above 48 kph when compared to probabilities based on CISS reconstructed delta-v. AACN algorithms are currently based on reconstructed delta-v but are triggered by EDR delta-v in the field. An analysis performed to determine the effect of the difference between the source of delta-v on the AACN notification threshold showed differences for AACN algorithms trained on EDR delta-v compared to reconstructed delta-v. The results of this study showed that the trigger threshold for AACN notification will differ by delta-v source, and this difference leads to variation in injury prediction.
Watson, Richard A., Bonugli, Enrique, Greenston, Mathew
Event Data Recorder - Retrieval Tool ProtocolJ1698/2_202307 (Current)7/14/2023
This SAE Recommended Practice utilizes existing industry standards to identify a common physical interface and define the protocols necessary to retrieve records stored by light duty vehicle event data recorders (EDRs). To accomplish this, the SAE J1962 diagnostic connector is designated as the primary physical interface for EDR retrieval tools. This SAE Recommended Practice is intended to be used for the development of EDR retrieval tools. Retrieval tools are intended to interface with light duty vehicles and to produce EDR record reports with data element formats specified by SAE J1698-1. Limitations: This SAE Recommend Practice specifies how EDR records should be imaged, translated, and reported by EDR retrieval tools. It does not specify how EDR records are recorded and stored by individual vehicles. This SAE Recommended Practice addresses EDR record retrieval (including imaging and translating) via the connection to the vehicle’s SAE J1962 connector or via direct connection to an electronic control unit (ECU) containing an EDR record. Direct connection to an ECU may require the use of specialized interface adapters.
Event Data Recorder Committee
TOC99-05-02-0TOC4/12/2023
TOC
Tobolski, Sue
Multi-Objective Bayesian Optimization Supported by Deep Gaussian Processes2023-01-00314/11/2023
A common scenario in engineering design is the evaluation of expensive black-box functions: simulation codes or physical experiments that require long evaluation times and/or significant resources, which results in lengthy and costly design cycles. In the last years, Bayesian optimization has emerged as an efficient alternative to solve expensive black-box function design problems. Bayesian optimization has two main components: a probabilistic surrogate model of the black-box function and an acquisition functions that drives the design process. Successful Bayesian optimization strategies are characterized by accurate surrogate models and well-balanced acquisition functions. The Gaussian process (GP) regression model is arguably the most popular surrogate model in Bayesian optimization due to its flexibility and mathematical tractability. GP regression models are defined by two elements: the mean and covariance functions. In some modeling scenarios, the prescription of proper mean and covariance functions can be a difficult task, e.g., when modeling non-stationary functions and heteroscedastic noise. Motivated by recent advancements in the deep learning community, this study explores the implementation of deep Gaussian processes (DGPs) as surrogate models for Bayesian optimization in order to build flexible predictive models from simple mean and covariance functions. The proposed methodology employs DGPs as the surrogate models and the Euclidean-based expected improvement as the acquisition function. This approach is compared with a strategy that employs GP regression models. These methodologies solve two analytical problems and one engineering problem: the design of sandwich composite armors for blast mitigation. The analytical problems involve non-convex and segmented Pareto fronts. The engineering problem involves expensive finite element simulations, three design variables, and two expensive black-box function objectives. The results show that the architecture of the DGP model plays an important role in the performance of the optimization approach. If the DGP architecture is adequate, the implementation of DGPs produces satisfactory results; otherwise, the use of GP regression models is preferable.
Valladares, Homero, Tovar, Andres
An Evaluation of the Automatic Emergency Braking and Forward Collision Warning System in a 2014 Subaru Forester2023-01-06214/11/2023
Testing was conducted to evaluate the performance of the 2014 Subaru Forester’s North American Generation 1 EyeSight system at speeds between 6 and 57 miles per hour (mph). The testing utilized a custom-built foam stationary vehicle target designed to withstand 60+ mph impact speeds. Testing measured the Time to Collision (TTC) values of the visual/audible component of the forward collision warning that was presented to the driver. In addition, the testing quantified the TTC and Time to Collision 2 (TTC2) response of the Automatic Emergency Braking (AEB) system including the timing and magnitude of the stage one braking response and the timing and magnitude of the stage two braking response. The results of the testing add higher speed Forward Collision Warning (FCW) and AEB testing scenarios to the database of publicly available tests from sources like the Insurance Institute for Highway Safety (IIHS), which currently evaluates vehicles’ AEB systems at speeds of 12 and 25 mph.
Harrington, Shawn, Martin, Nicholas
Engine Idle Creep Testing and Modeling of Vehicles Equipped with CVT, DCT, and Conventional Automatic Transmissions2023-01-06204/11/2023
Determining impact speeds is an important factor in any accident reconstruction. Event data recorders are now commonplace in on-road vehicles and provide an added tool for the accident reconstructionist. However, in low-speed collisions where impact severity is often important, event data recorders fail to record data as the minimum threshold for impact severity sometimes is not met. Alternatively, damage-based methods may be ineffective in quantifying the severity of the impact due to a lack of defined vehicle crush damage. These types of scenarios oftentimes present themselves as a bullet vehicle in the beginning processes of accelerating from a stop or when a stopped target vehicle is rear-ended from behind by the bullet vehicle. A specific subset of this scenario might entail the foot of the driver of the bullet vehicle coming off the brake pedal, allowing the bullet vehicle to “creep” forward at engine idle speeds and impacting the target vehicle resulting in no visible crush damage to either vehicle. Eighteen vehicles with conventional automatic transmissions were tested, which included sedans, sport utility vehicles (SUVs), pickup trucks, and vans. Two vehicles (one sedan, one wagon) equipped with dual-clutch transmissions (DCTs) and three vehicles (one sedan, one hatchback, and one wagon) equipped with continuously variable transmissions (CVTs) were also tested. These vehicles were allowed to accelerate at idle with the brake pedals released. Acceleration, speed, distance, and engine speed data were collected for multiple vehicles runs in both forward and reverse directions over level ground. The data resulting from this study were then compared/contrasted among the different drivetrains and also previously published literature to determine similarities and differences. Previous study data sets were also incorporated with the authors’ data to improve predicted vehicle speed.
Timbario, Thomas A., Stoner, Jacob, Sheldon II, Stuart
Fuel Flow Rates for Jet-Powered Commercial Aircraft Taxi OperationsAIR8035 (Current)4/4/2023
This SAE Aerospace Information Report (AIR) summarizes prior empirical findings (AIAA 2018-3991; Chati, 2018) to recommend a modified baseline fuel flow rate model for jet-powered commercial aircraft during taxi operations on the airport surface that better reflects operational values. Existing standard modeling approaches are found to significantly overestimate the taxi fuel flow rate; therefore, a modified multiplicative factor is recommended to be applied to these existing approaches to make them more accurate. Results from the analysis of operational flight data are reported, which form the basis for the modeling enhancements being recommended.
A-21 Aircraft Noise Measurement Aviation Emission Modeling
Event Data RecorderJ1698_202303 (Current)3/1/2023
This SAE Recommended Practice describes common definitions and operational elements of Event Data Recorders. The SAE J1698 series of documents consists of the following: SAE J1698-1 - Event Data Recorder - Output Data Definition: Provides common data output formats and definitions for a variety of data elements that may be useful for analyzing vehicle crash and crash-like events that meet specified trigger criteria. SAE J1698-2 - Event Data Recorder - Retrieval Tool Protocol: Utilizes existing industry standards to identify a common physical interface and define the protocols necessary to retrieve records stored by light duty vehicle Event Data Recorders (EDRs). SAE J1698-3 - Event Data Recorder - Compliance Assessment: Defines procedures that may be used to validate that relevant EDR output records conform with the reporting requirements specified in Part 563, Table 1 during the course of FMVSS-208, FMVSS-214, and other applicable vehicle level crash testing.
Event Data Recorder Committee
TOC99-05-01-0TOC2/28/2023
Tobolski, Sue
Automation of eddy current dynamometer for engine testing and calibration2022-36-00042/10/2023
The digital control of a dynamometer is essential for test of internal combustion engines, because it acts in command of the testing routine. This work consists in the PID (proportional, integrative and derivative) control of the Foucault current dynamometer, from the Institute of Technological Research - IPT. During the control development was made plant differents types, and differents controllers, that was validated by form experimental and theory. Digital controls were achieved using LabVIEW® and Simulink® software, which served to compare and simulate them, thereby making possible a validation. For the system in real-time, controlled by LabVIEW®, it was necessary to identify the dynamometer model, which corresponds to an eddy current brake, model I2D. The speed control developed by means of these tools is intended to adjust the speed of the dynamometer. In order to carry out the tests, this speed parameter must follow the desired values (set point) and maintain the relationship of electric current intensity with the throttle opening of the motor. In this way, the developed project opens up a perspective of greater reliability, that is, its responses are met in relation to the stabilization and transient of controlled values in tests performed with bench dynamometers. In the academic field, it ranges from theoretical to experimental, as it applies a scientific method to practical tests carried out in a validation instrument. Therefore, this contribution developed a digital control from the physical parameters of the project using the white box identification method, made by approximating a mathematical model of a direct current motor (DC). The black box method was also implemented, made by identifying the reaction of the dynamometer to a step response, in the experimental scope. The simulated reactions were compared with the experimental ones and showed equivalence between them.
da Mata, Valter Manuel Jardim, Moscardini, Demerson, da Silva Pereira, Bruno,  , Antônio, Maria Laganá, Armando
Event Data Recorder - Output Data DefinitionJ1698/1_202301 (Current)1/13/2023
This SAE Recommended Practice provides common data output formats and definitions for a variety of data elements that may be useful for analyzing vehicle crash and crash-like events that meet specified trigger criteria. The document is intended to govern data element definitions, to provide a minimum data element set, and to specify EDR record format as applicable for light-duty motor vehicle Original Equipment applications.
Event Data Recorder Committee
Comparison of Rear Impact Crash Reconstructions to Event Data Recorders in the Crash Investigation Sampling System Database2022-01-50699/9/2022
This article compares the results of automotive accident reconstructions to event data recorder (EDR) data from vehicles involved in rear-end collisions. Accident reconstructions in the Crash Investigation Sampling System (CISS) database calculate crash severity expressed as the impact-related change in velocity (delta-V) experienced by a vehicle. The accuracy of the CISS-reconstructed delta-V in rear impacts was assessed by comparison to the delta-V recorded during the crash by the EDR on board the rear-ended vehicles. The CISS database was searched for single rear impact cases with a CISS-reconstructed delta-V as well as an EDR download. A total of 256 cases met these criteria. On average, the CISS-reconstructed delta-V was 4.0% lower than the delta-V recorded by the EDR. The accuracy of the CISS reconstructions varied with crash configuration, vehicle type, collision partner, and crash severity. Crash severity had the largest effect on accuracy, with low-speed reconstructions overestimating the EDR delta-V by 36% on average.
Watson, Richard, Cormier, Joseph, Bonugli, Enrique, Greenston, Mathew
Hierarchical Bayesian Networks for Predicting the Structural Behavior of Coupled Systems2022-01-09426/15/2022
Simulations play an important role in the continuing effort to reduce development time and risks. However, large and complex models are necessary to accurately simulate the dynamic behavior of complex engineering systems. In recent years, the use of data-driven models based on machine learning (ML) algorithms has become popular for predicting the structural dynamic behavior of mechanical systems. Due to their advantages in capturing non-linear behavior and efficient calculation, data-driven models are used in a variety of fields like uncertainty quantification, optimization problems, and structural health monitoring. However, the black box structure of ML models reduces the interpretability of the results and complicates the decision-making process. Hierarchical Bayesian Networks (HBNs) offer a framework to combine expert knowledge with the advantages of ML algorithms. In general, Bayesian Networks (BNs) allow connecting inputs, parameters, outputs, and experimental data of various models to predict the overall system-level dynamic behavior. This characteristic of BNs enables a divide and conquer approach. Hence, complex engineering systems can be split into more easily describable subsystems. HBNs are an extension of BNs that can use knowledge about the structure of the data to introduce a bias that can contribute to improving the modelling result. In this work, an approach to design a HBN is presented where each model in the network can be a parametric reduced finite-element models. The influence of the hierarchical approach is evaluated by comparing a HBN and a BN of the model from the Sandia structural dynamics challenge.
Hülsebrock, Moritz, Schmidt, Hendrik, Stoll, Georg, Atzrodt, Heiko
Cheeseman Award Paper: Damper Model Identification Using Hybrid Physical and Machine Learning Based ApproachF-0078-2022-13225/10/2022
In this paper the identification of a time domain model of a helicopter main rotor lead-lag damper is discussed. Previous studies have shown that lead-lag dampers have a significant contribution to the overall aircraft dynamics, therefore an accurate damper model is essential to predict complex phenomena such as, instabilities, limit cycles, etc. Due to the inherently nonlinear dynamics and the complex internal architecture of these components, the model identification can be a challenging task. In this paper, a hybrid physical/machine learning based approach, has been used to identify a damper model based on experimental test data. The model, called grey box, consists of a combination of a white box, i.e. a physical model described by differential equations and a black box, i.e. regression numerical model. The white box approximates the core physical behaviour of the damper while the black box improves the overall accuracy by capturing the complex dynamic not included in the white box. The paper shows that, at room temperature, the grey box is able to predict the damper force when either a multi-frequency harmonic or a random input displacement is imposed. The model is validated up to 20 Hz and for the entire damper dynamic stroke.
Zilletti, Michele, Fosco, Ermanno
The Future of Managing Mission-Critical SystemsTBMG-454914/1/2022
There’s no doubt the complexity of aerospace design systems is constantly increasing, driven by new demands on architecture and next-generation technologies. As a result, the costs and time associated with the creation, certification and deployment of mission-critical electronics hugely heighten if the systems are not managed in a new way.
Study on Estimation of Traveling Trajectory Using the Recording Data in the Event Data Recorder2022-01-08373/29/2022
In order to investigate traffic accidents and determine their causes, first it is necessary to clarify the circumstances in which they occurred. The traveling trajectory of the vehicle(s) involved prior to the collision is an important part of such clarification. In this study, we conducted experiments on a vehicle with an event data recorder (EDR) and examined its pre-collision trajectory estimated from data recorded by EDR, aiming to obtain such trajectories based on quantitative data recorded by EDRs. In the experiment, the test vehicle with an EDR had also a high-precision measuring system onboard that determined the vehicle’s position by a global positioning system (GPS) and measured vehicle behavior. The vehicle was driven on a test course, with the EDR and the measuring system recording their data simultaneously. The vehicle speed and yaw rate data recorded by the EDR were integrated to get an estimated trajectory of the vehicle. The comparison of the EDR-based traveling trajectory with the GPS positioning results revealed generally similar results when the vehicle behavior changed slowly (rounding curves, etc.). When the vehicle behavior changed sharply (zipping through a slalom, etc.), the difference between the EDR-based traveling trajectory and the GPS positioning results was larger than when it changed slowly. When the vehicle behavior changed sharply, the EDR-based traveling trajectory could not be fully reproduced because the sampling frequency of the EDR-based vehicle speed and yaw rate was lower. The results of this experiment suggest that the traveling trajectory can be estimated from the EDR-based vehicle speed and yaw rate if the vehicle behavior changes slowly. The results also indicated that, to get the traveling trajectory while the vehicle behavior keeps changing rapidly, we need to make the EDR sampling frequency higher than the current ones.
Matsumura, Hideki, Itoh, Tatsuya
Physics-Guided Sparse Identification of Nonlinear Dynamics for Prediction of Vehicle Cabin Occupant Thermal Comfort2022-01-01593/29/2022
Thermal cabin comfort is the largest consumer of battery energy second only to propulsion in Battery Electric Vehicles (BEV’s). Accurate prediction of thermal comfort in the vehicle cabin with fast turnaround times will allow engineers to study the impact of various thermal comfort technologies and develop energy efficient Heating, Ventilation and Air Conditioning (HVAC) systems. In this study a novel data-driven model based on physics-guided Sparse Identification of Nonlinear Dynamics (SINDy) method was developed to predict Equivalent Homogeneous Temperature (EHT), Mean Radiant Temperature (MRT) and cabin air temperature under transient conditions and drive cycles. EHT is a recognized measure of the total heat loss from the human body that can be used to characterize highly non-uniform thermal environments such as a vehicle cabin. The SINDy model was trained on drive cycle data from Climatic Wind Tunnel (CWT) for a representative Battery Electric Vehicle. The performance of the trained model was evaluated on drive cycle data from a Coldbox test for the same vehicle. Based on a physics-based transient lumped cabin comfort model three state variables (EHT, MRT and cabin air temperature) and four control inputs were chosen to develop the SINDy model. In addition to the control inputs, interaction terms between the different control inputs were also included. Despite training with limited data (two trajectories of different time lengths) the model learned the dynamics of the vehicle cabin and was able to predict EHT, MRT and cabin air temperature for a new set of control inputs reasonably well. The SINDy model can be used for predictions of thermal comfort under transient conditions with significantly lower turnaround times compared to physics-based simulations. It is an alternative to black box models by providing insights into the mechanisms driving a dynamical system.
Warey, Alok, Kaushik, Shailendra, Han, Taeyoung
Identifying Pedal Misapplication Behavior Using Event Data Recorders2022-01-08173/29/2022
Pedal misapplication (PM) crashes, i.e., crashes caused by a driver pressing one pedal while intending to press another pedal, have historically been identified by searching unstructured crash narratives for keywords and verified via labor-intensive manual inspection. This study proposes an alternative method to identify PM crashes using event data recorders (EDRs). Since drivers in emergency braking situations are motivated to hit the brake hard, it follows that drivers in emergency braking situations that commit a PM would likewise hit the accelerator hard, likely harder than accelerator pedal application during normal driving. Thus, the time-series accelerator pedal position and the derived accelerator pedal application rate were used to isolate accelerator misapplications. Additional strategic filters were applied based on characteristics observed from previous PM analyses to reduce false positive PM identifications. These include a crash type filter, since PM crashes have been shown to manifest as majority road departure, end departure, rear-end, and forward impact crash types. After analyzing pre-crash EDR data from the National Automotive Sampling System Crashworthiness Data System (NASS/CDS) case years 1997 to 2015, evidence of PM was observed in 4.3% of weighted events. This result was substantially higher than the previously estimated 0.2% PM frequency [1,2]. The time-to-collision (TTC) at the point of PM was calculated for each case, and over 50% of cases had a TTC of less than 2.0 seconds. Over one-third of these drivers engaged the accelerator to 99% of pedal stroke or above and over one-eighth of drivers engaged both the brake and the accelerator pedals simultaneously during the recorded pre-crash period.
Smith, Colin P, Sherony, Rini, Gabler, H. Clay, Riexinger, Luke E
Rear-End Impacts - Part 1: Field and Test Data Analysis of Crash Characteristics2022-01-08593/29/2022
Prior to developing or modifying the protocol of a performance evaluation test, it is important to identify field relevant conditions. The objective of this study was to assess the distribution of selected crash variables from rear crash field collisions involving modern vehicles. The number of exposed and serious-to-fatally injured non-ejected occupants was determined in 2008+ model year (MY) vehicles using the NASS-CDS and CISS databases. Selected crash variables were assessed for rear crashes, including severity (delta V), impact location, struck vehicle type, and striking objects. In addition, 15 EDRs were collected from 2017 to 2019 CISS cases involving 2008+ MY light vehicles with a rear delta V ranging from 32 to 48 km/h. Ten rear crash tests were also investigated to identify pulse characteristics in rear crashes. The tests included five vehicle-to-vehicle crash tests and five FMVSS 301R barrier tests matching the struck vehicle. The analysis of NASS-CDS and CISS data indicates that more than 50% of exposed (MAIS 0+F) occupants were involved in a rear crash with a distributed impact and with no over-ride or under-ride. More than 93% were with a 6 o’clock PDOF. The average rear delta V was 21.3 ± 1.2 km/h (median 18.2, 90th CI 9.7-33.4) in the exposed (MAIS 0+F) occupant sample. For serious-to-fatally (MAIS 3+F) injured occupants, more than 88% were in a fully distributed rear impact and in 6 o’clock PDOF; 10.7% were in offset impacts and less than 1% were in narrow impacts. The average delta V was 35.9 ± 5.2 km/h (median 24.4, 90th CI 16.2-58.3) for serious-to-fatally (MAIS 3+F) injured occupants. The results from the field data suggest that rear impacts with a delta V of 34-38 km/h and a 6 o’clock PDOF are representative of the average rear-crash scenario causing serious-to-fatal injury. Compared to the exposed occupants in the field data analysis, a 40 km/h rear sled is more than 1.9-times more severe than the average rear impact crash severity in terms of delta V and 4.4-times more severe in terms of crash energy. It is 1.11-times more severe than the average delta V for the serious-to-fatally (MAIS 3+F) injured occupant. It accounts for more than 90% of all occupants with injury (MAIS 1+F). EDR data was helpful in accessing overall delta-V and impact duration information. For example, the average delta V in the EDRs included in the analysis was 38.7 ± 7.6 km/h and the average pulse duration was 129.6 ± 24.3 ms. Instrumentation data from crash tests provided more refined information for a detailed analysis of pulse characteristics. The acceleration data obtained from the five vehicle-to-vehicle crash tests was investigated and compared to matched FMVSS 301 tests with the same struck vehicle. The data was scaled to a 40 km/h delta V pulse for comparison purposes. Bi-modal pulse shape characteristics were observed. The peak acceleration averaged 18.4 g in the vehicle-to-vehicle tests and 21.4 g in the FMVSS 301R tests. The average pulse duration was 147.2 ms and 123.8 respectively. The results from this study suggest that using a 34-40 km/h bimodal-rear pulse with a 130-160 ms duration is representative of a serious real world rear impact.
Parenteau, Chantal, White, Samuel, Burnett, Roger, Stephens, Gregory, Michalski, David
Validation of Frozen Speedometer Readings Against EDR Data2022-01-08213/29/2022
Around the turn of this century, the automotive industry introduced a new type of technology to drive the gauges on a vehicle’s instrument cluster. The change was unannounced to the collision reconstruction world, but soon after, investigators observed a marked increase in crashed vehicles displaying frozen gauges at what often appeared to be correct readings. The new technology was the use of stepper motors which require power to return to the zero position. Hence if electrical power is lost, the gauges stop in position. There have been a number of previous papers covering the operation of the instruments and crash testing of cars and motorcycles to establish the ability of the instruments to withstand the forces on the instrument during a collision. This paper aims to compare the frozen instrument readings from real world collisions with the available EDR data from the crashed vehicles. With the assistance of the collision reconstruction community, a large dataset of 236 vehicles with frozen speedometer readings were compared with EDR and other corroborating methods. This paper reviews the current state of knowledge, compares the instrument readings of each of the 236 vehicles against the available EDR data or other corroborating method. It then assesses each case against the criteria proposed by Goddard and Price [3, 4] to assess if the cases that are being filtered out for selection, are accurate readings. It was found the existing criteria was an effective filter in removing the majority of cases capable of producing erroneous readings. However, it was found that some erroneous readings were present in low speed impacts. With the addition of a minimum speed criteria, the likelihood of an erroneous reading was greatly reduced. At recorded speedometer readings over 80 Km/h, the corroboration with the EDR speed had a standard deviation of 4%.
Goddard, Christopher H., Anderson, Steve
TOC99-04-01-0TOC3/18/2022
Tobolski, Sue
Items per page:
1 – 50 of 309