Road roughness is the primary source of vehicle vibrations. This thesis investigates model-based methods for estimating road roughness in terms of the International Roughness Index (IRI) by measuring the chassis vibrations of the vehicle. This can provide NIRA Dynamics AB with a cost-effective pavement monitoring solution. Initially, system identification is performed on a physical car to estimate model parameters that reflect reality. Subsequently, two model-based IRI estimation methods are developed. One method relies on a transfer function between vertical chassis vibrations and the IRI according to a quarter-car model. The second method aims first to estimate the longitudinal road profile using a Kalman filter, and then calculate the IRI values from the estimated profile. This method can be implemented computationally efficiently and also offers the possibility of estimating the IRI using lateral vibrations. Both methods are validated using real-world data, and their performance is similar when using vertical vibrations, with the IRI estimation error’s standard deviation being roughly 10% to 20% of the reference value. However, the results are considerably worse when the estimation is purely based on lateral vibrations, indicating that lateral vibrations are not feasible for model-based IRI estimation, and the reasons for this are discussed.

Als elastische Koppelglieder üben Elastomere im Fahrwerk maßgeblichen Einfluss auf die Elastokinematik der Radaufhängung und damit auf die Radstellgrößen aus. Bereits einfache zylindrische Probekörper zeigten in der Vergangenheit Veränderlichkeit der Federsteifigkeiten bei mehrachsiger Belastung auf. Daraus ließ sich ableiten, dass das Bauteilverhalten von zylindrischen Elastomerkörpern, bzw. Fahrwerklagern, schwer vorhersagbar ist und dieser Effekt durch die Verwendung komplexer Geometrien zusätzlich beeinflusst wird. Eigens im Rahmen dieser Arbeit entwickelte Prüfstände ermöglichten die Untersuchung von statischen, mehrachsigen Belastungen sowie zyklischen, einachsigen Belastungen zweier Fahrwerklager einer typischen Vorderradaufhängung, dem Mc-Pherson-Federbein. Für beide Lagertypen wurden derart signifikante Federsteifigkeitsänderungen unter mehrachsiger Belastung beobachtet, dass ein Einfluss auf die Elastokinematik sicher schien. Unter einachsiger, zyklischer Belastung bestätige sich, dass einfache rheologische Ersatzmodelle, wie der Ansatz nach Kelvin-Voigt für eine Modellierung des viskoelastischen Materialverhaltens weiterhin Gültigkeit besitzen. Ein in einem Mehrkörpersimulationsprogramm erstelltes Gesamtfahrzeugmodell zeigte zunächst für die Auslegung der Prüfstände relevante Beanspruchungen in den Fahrwerklagern auf und diente weiterhin als Grundlage für die Untersuchung des Einflusses von Wechselwirkungen auf die Radstellgrößen. Dazu behandelt die Arbeit die Modellierung eines neuen Fahrwerklagermodells innerhalb dieser MKS-Umgebung. Dieses Modell bietet dem Nutzer die Möglichkeit für eine Abbildung von Wechselwirkungen an zylindrischen, zyklisch beanspruchten Bauteilen. Ein spezielles Fahrmanöver zeigte den Einfluss der Wechselwirkungen auf die Radstellgrößen auf und zeigte, dass die Berücksichtigung von Wechselwirkungen für eine Fahrwerkauslegung relevant ist. Darüber hinaus wurde mit kommerziell verfügbaren Materialmodellen für die Beschreibung es Materialverhaltens von Elastomeren innerhalb der Finite-Elemente-Methode aufgezeigt, dass bei einachsiger Belastung die Federsteifigkeiten der Fahrwerklager auf Grundlage der in der Literatur üblichen Auslegungsgleichungen näherungsweise abgebildet werden können, jedoch für mehrachsige Spannungszustände keine hinreichende Aussage über das Bauteilverhalten getroffen werden kann. Hier wurde weiterer Forschung- bzw. Erweiterungsbedarf an den FE-Programmen aufgezeigt.

Highway roughness is a concern for both the motoring public and highway authorities. Roughness may even increase the risk of crashes. Rail-highway grade crossings are particularly problematic. Roughness may be due to deterioration or simply due to the way the crossing was built to accommodate grade change, local utilities, or rail elevation. With over 216,000 crossings in the US, maintenance is a vast undertaking. While methods are available to quantify highway roughness, no method exists to quantitatively assess the condition of rail crossings. Conventional inspection relies on a labor-intensive process of qualitative judgment. A quantifiable, objective and extensible procedure for rating and prioritizing improvement of crossings is thus desired. In this dissertation, a 3D infrastructure condition assessment model is developed for evaluating the condition and performance of rail highway grade crossings. Various scanning techniques and devices are developed or used to obtain the 3D “point cloud” or surface as the first step towards quantifying crossing roughness. Next, a technique for repeatable field measurement of acceleration is presented and tested to provide a condition index. Acceleration-based metrics are developed, and these can be used to rate and compare crossings for improvement programs to mitigate potential vehicle damage and provide passenger comfort. A vehicle dynamic model is next customized to use surface models to estimate vertical accelerations eliminating the need for field data collection. Following, crossing roughness and rideability is estimated directly from 3D point clouds. This allows isolation of acceleration components derived from the surface condition and original design profile. Finally, a practice ready application of the 3D point cloud is developed and presented to address hump crossing safety. In conclusion, the dissertation presents several methods to assess the condition and performance of rail crossings. It provides quantitative metrics that can be used to evaluate designs and construction methods, and efficiently implement cost effective improvement programs. The metrics provide a technique to measure and monitor system assets over time, and can be extended to other infrastructure components such as pavements and bridges.

Le contrôle de stabilité est un sujet essentiel dans les systèmes avancés d’aide à la conduite développés par les constructeurs et équipementiers automobiles. Les systèmes de sécurité actifs sont devenus un standard dans les véhicules particuliers, tels que : le contrôle électronique de la stabilité (ESC) et le système de contrôle de traction (TCS). La description du comportement dynamique du véhicule pendant le mouvement, est fondamental dans le fonctionnement des nouveaux systèmes de sécurité active. Certains systèmes actifs sont déjà implémentés dans des véhicules standards comme des options supplémentaires, pour améliorer la sécurité sur la route ou pour le confort du conducteur et des passagers. Cependant, ces systèmes ont besoin d’informations sur la dynamique de véhicule, qui représente les caractéristiques de mouvement du véhicule sur la route. L’accès à ces informations est souvent difficile, pour des raisons technologiques ou économiques. De ce fait, nous développons des algorithmes, basés sur la technique d’observation d’état, pour estimer une partie de ces variables notamment, les efforts dynamiques du contact pneumatique/chaussée et l’angle de dérive dans son environnement. En revanche, ces systèmes sont conçus pour faire face à l’état actuel du véhicule où la situation de danger a toujours eu lieu, la capacité de ces systèmes est limitée à minimiser les effets de danger. L’objectif ultime est de prévoir et d’éviter efficacement un accident avant qu’il se produise. Par conséquent, ce travail est dédié aussi à développer une méthode de prédiction des risques pour rappeler au conducteur la vitesse de sécurité pour négocier les virages à venir. Dans un premier temps, nous développons dans ce mémoire une nouvelle approche pour estimer la répartition de la charge verticale sur chaque roue dans un environnement réel. L’influence de l’angle de pente est considérée dans la phase de reconstruction du modèle du véhicule. Les forces verticales sont estimées en utilisant un filtre de Kalman. Afin d’estimer la force latérale du pneu, un filtre de Kalman entendu et un filtre Particulaire ont appliqués pour tenir compte des non-linéarités du modèle de véhicule. Deux techniques différentes d’observateurs sont proposées et comparées avec des données expérimentales. Dans un deuxième temps, nous étendons, à l’instant futur, la prise en compte de l’évaluation de risque d’accidents. La prédiction des paramètres de la dynamique du véhicule, l’évaluation du risque potentiel ainsi que la détermination d’une vitesse d’alerte à l’approche des virages, sont introduites pour réduire le risque potentiel d’accident dans les virages. Enfin, la dernière partie du mémoire est consacrée à l’application en temps réel, sur un véhicule démonstrateur, du processus d’observation d’état développé précédemment. Les résultats expérimentaux sont réalisés pour démontrer la performance des estimateurs intégrés en temps réel. / Nowadays, a variety of advanced driving assistance systems are being developed by research centers and automobile manufactures. Stability control is an essential topic in the modern industrial automobile society. Driving safety is widely concerned in the passenger cars to prevent potential risks. More and more electronic active safety systems are fitted out as a standard option, such as Electronic Stability Control (ESC) and Traction Control System (TCS). These safety systems are efficient in helping the driver maintain control of the car and also are considered highly cost-effective. However, for the future development trend of these systems, a more complex and integrated control unit requires more information about the vehicle dynamics. Some fundamental parameters such as tire road forces and sideslip angle are effective in describing vehicle dynamics. Nevertheless, it is lacking an effective and low-cost sensor to measure directly. Therefore, this study presents amethod to estimate these parameters using observer technologies and low-cost sensors which are available on the passenger cars in real environment. In addition, these systems are designed for dealing with vehicle current state where danger situation has always occurred, the capacity of these systems is limited to minimize the effects. We were wondering whether shall we predict and effectively avoid a crash before it occurred. Therefore, this work is also addressed to develop a risk prediction method for proposing driver a safe speed to negotiate the upcoming curves. First, this dissertation develops a new approach to estimate the vertical load distribution in real environment. The influence of bank angle is considered in the phase of reconstruction of vehicle model. The vertical tire force on banked road is estimated by using Kalman filter. In order to estimate the lateral tire force, two nonlinear observers are addressed to solve the nonlinearity of vehicle model. The Extended Kalman filter is widely discussed in the previous literature, while we firstly use a Particle filter to estimate the vehicle dynamics parameters. Two different observer technologies are proposed and compared using the experimental data. Second, extending the consideration of road safety to the future instant. Prediction of vehicle dynamics parameters, evaluation of potential risk as well as establishment of advisory speed on curves are introduced to reduce the possibility of crash occurrence on curves. Last but not least, the real-time sampling and process system is presented, the estimator with EKF and PF has been developed as a real-time application. Experimental results are performed to demonstrate the performance of these integrated systems in real-time.

A aplicação do método de elementos finitos como ferramenta de análise vem crescendo na indústria ao longo dos anos, sendo que uma das áreas em desenvolvimento corresponde ao estudo de impacto e avaliação de resistência ao choque. O objetivo deste trabalho é avaliar a eficiência da defensa metálica maleável simples utilizada nas rodovias do Brasil, no que diz respeito à capacidade de conter um veículo desgovernado de 2000 quilogramas a 100 quilômetros por hora, através da simulação numérica pelo método dos elementos finitos. Para possibilitar a execução das análises, são realizados ensaios experimentais de corpos de prova retirados de um dos componentes da defensa metálica para a caracterização do material. Além disso, são desenvolvidas simplificações no modelo para permitir a realização de análises com baixo custo computacional. A partir do modelo elaborado, são avaliados os comportamentos de impacto na defensa metálica maleável simples em diversas condições de instalação encontradas em vias brasileiras com grande volume de veículos, incluindo-se condições inadequadas. As avaliações indicam que a defensa é incapaz de reter o veículo, ocorrendo o rompimento dos parafusos que unem as lâminas. As defensas em condições irregulares de manutenção apresentam resultados mais catastróficos, colocando mais risco aos ocupantes e pedestres. Por fim, é recomendada a elevação da classe do parafuso para 5.8, permitindo que a defensa retenha o veículo do teste 3-11 do NCHRP Report 350. / The application of Finite Element Method as an analysis tool in industry has been increasing over the years, so that one of the development areas corresponds to the study of impact and crashworthiness. The objective of this work is to assess the efficiency of a guardrail model used on Brazilian roads, with respect to the ability of containing a runaway vehicle of 2 ton at 100 miles per hour, through numerical simulation using finite element analysis. Experimental tests of samples taken from a component of the guardrail are performed to enable material characterization for the analysis. Furthermore, model simplifications are developed to permit the analysis with low computational cost. The impact behaviors are evaluated in various guardrail installation conditions found on high movement roads, including some inadequate ones. The analysis show that the guardrail is unable to retain the vehicle, shearing the W-beam splice bolts. The irregular maintenance condition guardrails present more catastrophic results, endangering even more the occupants and the pedestrians. Finally, it is recommended that the bolts are upgraded to 5.8, enabling the vehicle retention under test 3-11 of NCHRP Report 350.

Apresenta um modelo multicorpos completo de um veículo comercial leve considerando a flexibilidade do quadro para estudos de dirigibilidade. O modelo de multicorpos foi desenvolvido utilizando o software MSC.ADAMS e inclui a suspensão primária, a cabine, o sistema de esterçamento, o trem de força, o modelo do pneu e um quadro flexível. O comportamento direcional, a dirigibilidade e a interação veículo/pavimento são os fatores que definem o desempenho do veículo em relação à dinâmica lateral. Os parâmetros mais importantes considerando o comportamento lateral do veículo são: a velocidade longitudinal, a rigidez em curva do pneu, a elasticidade da suspensão e do sistema de esterçamento, a transferência lateral de carga, centro de rolamento e propriedades inerciais. Conseqüentemente, considerando a transferência lateral de carga, o uso do quadro flexível se mostra importante. O modelo completo foi validado com resultados experimentais. Típicas manobras para estudo de dirigibilidade foram simuladas, entre elas: curva de raio e velocidade constantes e manobras de dupla mudança de pista. Os resultados obtidos foram a aceleração lateral, a velocidade de guinada, o ângulo de deriva da trajetória, o ângulo de rolamento e o gradiente de esterçamento. Com a análise dos resultados, o veículo mostrou comportamentos diferentes em função da aceleração lateral. Em baixas acelerações laterais, o veículo apresenta um comportamento levemente sub-esterçante. Contudo, em acelerações laterais altas, o veículo se torna sobre-esterçante. Finalmente, uma análise no domínio da freqüência foi realizada e ambos modelos mostraram as mesmas características. / This dissertation presents a complete multibody model of a light commercial truck considering the frame flexibility for handling characteristics studies. The flexible multibody model was developed using MSC.ADAMS, and it includes a complete primary suspension, the cabin, the steering system, the powertrain, a tire model (Delft) and the flexible frame. The directional response behavior, the driveability and the vehicle/road interaction are the factors that define vehicle handling performance. The most important parameters concerning lateral behavior are: longitudinal velocity, tire cornering stiffness, suspension system elasticity, steering systems elasticity, lateral load transfer, roll center and inertial properties. Consequently, regarding lateral load transfer, the use of a flexible frame is very important. The complete model was validated with experimental results and for the purpose of vehicle analyses, typical standard handling maneuvers were undertaken including constant radius turn, constant velocity turn and double lane change. The results obtained were lateral acceleration, yaw rate, side slip angle, roll angle and the understeer gradient. Analyzing the results, the vehicle showed a changing behavior concerning steer. At low velocities and lateral accelerations, the vehicle is slightly understeer. However, at higher velocities and lateral accelerations, it becomes oversteer. Finally, a frequency domain analysis was undertaken using the simplified and the complete model and both models shown the same characteristics.

O escopo do trabalho é a modelagem de um coxim elastomérico de motor, elemento que possui papel fundamental no isolamento vibracional do conjunto motor e transmissão. O estudo do comportamento mecânico de materiais elastoméricos é um campo que é desenvolvido a muitos anos devida complexidade e não-linearidade desses componentes; e modelos matemáticos fiéis são diferenciais competitivos. Assim essa dissertação tem como objetivo principal propor a modelagem de um coxim de motor elastomérico utilizando a técnica de redes neurais para generalizar a função de transferência entre o deslocamento do coxim e a aceleração vertical de motor. Duas abordagens de treinamento - dados experimentais coletados em bancada hidráulica uniaxial e dados experimentais coletados em campo - são apresentadas com o intuito de identificar a condição que a rede neural apresenta melhor performance de generalização. Para tal comparação uma métrica baseada em área da densidade espectral de potência é apresentada para quantificar o desempenho do modelo na faixa de frequência estudada (0-40 Hz). Finalmente é realizada uma comparação com um modelo mecânico composto por molas e amortecedores combinados. Os resultados demonstram que tanto o treinamento realizado com dados de bancada quanto o modelo mecânico apresentam boa correlação de 0 Hz a 14 Hz enquanto o treinamento realizado com dados de campo tem boa correlação de 0 Hz a 14 Hz e de 23 Hz a 40 Hz uma vez que essa abordagem tem capacidade de capturar a histerese e parte da não linearidade da borracha. É demonstrado que na faixa de frequência na qual todos modelos não foram capazes de generalizar existe modo de vibrar onde um nó no coxim estudado que dificulta a modelagem. Conclui-se que a técnica de redes neurais possui grande potencial em sua utilização, apresentando resultados bastante satisfatórios, além de outras vantagens, como a velocidade de processamento da rede treinada. / This dissertation scope is the modeling of an elastomeric mount, element which has primordial role on powertrain vibration isolation. The study of the mechanic behavior of those elastomeric materials is a field being studied for several years agor since its complexity and non-linarity of those components; and a reliable math models are competitive edge. So this dissertation has as main object purpose the modeling of an elastomeric engine mount using neural networks technician to predict the transfer function between the mount displacement and engine vertical accelerationTwo training approachs experimental data gathered in a uniaxial hydraulic bench and experimental data gathered in field are presented with the objective of identify the condition which the neural network presents better prediction performance. So this comparison being made a metric based on power density spectral area is presented to quantify the performance on the frequency range studied (0-40 Hz). Finally a comparison is made with a mechanical model composed per springs and dumpers combined. The results show that even the training made with the bench data and the mechnical model present good correlation from 0 Hz to 14 Hz while the training made with field data has good correlation from 0 Hz to 14 Hz and 23 Hz to 40 Hz once this approach has the ability to capture the hysteresis and part of rubber non-linearity. It is demonstrated that the frequency range where all the models are not capable to predict, there is a vibration mode where there is a node on the studied mount which dificults the modeling. The conclusion is that the neural network technique has great potential on its usage, presenting very satisfactory results, among other vantages as the processing speed after the network is trained.

O comportamento da dinâmica lateral de veículos automotivos com esterçamento nas quatro rodas é estudado com o auxílio de uma ferramenta computacional de modelagem e simulação de sistemas multicorpos. São utilizadas duas estratégias de controle do esterçamento das rodas traseiras para a avaliação de suas principais características. Uma das estratégias esterça as rodas traseiras em função do ângulo de esterçamento do volante, e a outra esterça as rodas traseiras em função do ângulo de esterçamento do volante e da velocidade longitudinal do veículo. O teste utilizado para as análises é a entrada degrau de esterçamento, onde é avaliada a resposta transitória e de regime da aceleração lateral, da velocidade de guinada e do ângulo de rolagem dos veículos. Os resultados das simulações mostram que os veículos com esterçamento nas quatro rodas têm o tempo de resposta da velocidade de guinada maior, e da aceleração lateral e do ângulo de rolagem menores que o veículo convencional. Também apresentam uma diminuição no ganho da aceleração lateral, da velocidade de guinada e do ângulo de rolagem. Além disso apresentam uma característica subesterçante mais acentuada. Essas características são mais evidentes no veículo com estratégia de controle função do esterçamento do volante. Os resultados obtidos correspondem ao comportamento dos veículos com esterçamento nas quatro rodas apresentado na literatura. / The lateral dynamic behavior of the automotive vehicles is studied with the aid of a computational tool for multibody systems modeling and simulation. Two rear wheel steering control strategies are used for evaluation of the main four wheel steering characteristics. One strategy steer the rear wheels as a function of the steering wheel angle, and the other one, steer the rear wheels as a function of the steering wheel angle and the speed. The steer step input is the test used, where the lateral acceleration, the yaw rate and the roll angle transient and stead state response are evaluated. The simulation results show that the four wheel steering vehicles have a slower yaw rate time response, and a fast lateral acceleration and roll angle time response than the conventional vehicle. Also four wheel vehicles show a reduction in lateral acceleration, yaw rate a.nd roll angle gain. Moreover they are more understeer than the conventional vehicle. These characteristics are particularly more evident on the vehicle with steer dependent system. The obtained results correspond with four wheel steering vehicles behavior founded in literature.

Este trabalho apresenta o estudo da dinâmica vertical de um veículo off-road de competição e do comportamento elastocinemático de suas suspensões primárias de duplo estágio com o auxílio da ferramenta computacional de simulação de sistemas multicorpos MSC-ADAMS. O modelo multicorpos do veículo inclui os modelos da suspensão dianteira, suspensão traseira, sistema de direção, pneus e massa suspensa. As análises elastocinemáticas das suspensões definiram geometrias que permitem longos cursos de trabalho das suspensões sem variações indesejadas de parâmetros de projeto. Com a análise modal do modelo de veículo completo foram definidos os valores de rigidez das molas das suspensões de duplo estágio que resultaram em valores de freqüência natural no primeiro estágio semelhantes aos indicados para carros de passeio e no segundo estágio, próximos dos encontrados em veículos de competição. A análise de conforto do veículo durante simulação de passagem por trecho de pista demonstrou que os elementos de força definidos na análise modal resultam em níveis de conforto raramente encontrados em veículos de competição. / This dissertation presents the study of a off-road competition vehicle ride and the double rate suspensions elastokinematic behavior using the multibody software MSC-ADAMS. The vehicle multibody model includes the rear and front suspensions, the steering system, the tires and the sprung mass. The suspensions elastokinematic analysis defined geometric configurations that allowed long jounce and rebound travel of wheel, without undesirable project parameters variations. In the modal analysis of the vehicle multibody model, the springs rates that results in natural frequencies values similar to passenger cars for the first suspension stage and, for the second stage, similar to competition vehicles, were defined. The ride analysis during the simulation of the vehicle running on a rough track showed that the force elements defined on the modal analysis resulted in a good ride quality, rarely found in competition vehicles.

La voiture autonome pourrait réduire le nombre de morts et de blessés sur les routes tout en améliorant l'efficacité du trafic. Cependant, afin d'assurer leur déploiement en masse sur les routes ouvertes au public, leur sécurité doit être garantie en toutes circonstances. Cette thèse traite de l'architecture de planification et de contrôle pour la conduite automatisée et défend l'idée que l'intention du véhicule doit correspondre aux actions réalisées afin de garantir la sécurité à tout moment. Pour cela, la faisabilité cinématique et dynamique de la trajectoire de référence doit être assurée. Sinon, le contrôleur, aveugle aux obstacles, n'est pas capable de la suivre, entraînant un danger pour la voiture elle-même et les autres usagers de la route. L'architecture proposée repose sur la commande à modèle prédictif fondée sur un modèle bicyclette cinématique afin de planifier des trajectoires de référence sûres. La faisabilité de la trajectoire de référence est assurée en ajoutant une contrainte dynamique sur l'angle au volant, contrainte issue de ces travaux, afin d'assurer que le modèle bicyclette cinématique reste valide. Plusieurs contrôleurs à haute-fréquence sont ensuite comparés afin de souligner leurs avantages et inconvénients. Enfin, quelques résultats préliminaires sur les contrôleurs à base de commande sans modèle et leur application au contrôle automobile sont présentés. En particulier, une méthode efficace pour ajuster les paramètres est proposée et implémentée avec succès sur la voiture expérimentale de l'ENSIAME en partenariat avec le laboratoire LAMIH de Valenciennes. / Autonomous vehicles are believed to reduce the number of deaths and casualties on the roads while improving the traffic efficiency. However, before their mass deployment on open public roads, their safety must be guaranteed at all time.Therefore, this thesis deals with the motion planning and control architecture for autonomous vehicles and claims that the intention of the vehicle must match with its actual actions. For that purpose, the kinematic and dynamic feasibility of the reference trajectory should be ensured. Otherwise, the controller which is blind to obstacles is unable to track it, setting the ego-vehicle and other traffic participants in jeopardy. The proposed architecture uses Model Predictive Control based on a kinematic bicycle model for planning safe reference trajectories. Its feasibility is ensured by adding a dynamic constraint on the steering angle which has been derived in this work in order to ensure the validity of the kinematic bicycle model. Several high-frequency controllers are then compared and their assets and drawbacks are highlighted. Finally, some preliminary work on model-free controllers and their application to automotive control are presented. In particular, an efficient tuning method is proposed and implemented successfully on the experimental vehicle of ENSIAME in collaboration with the laboratory LAMIH of Valenciennes.