In this dissertation, a study on the real-time dynamic simulation of multibody vehicle systems using suspension composite joints is presented. The suspension composite joints are derived and utilized to reduce the computation time of simulation without any degradation of kinematical accuracy of suspension systems. The joints are modeled using massless links on the suspension members that have small masses but have important kinematic functions, and the kinematics of knuckles or axles of the suspension systems are derived. Typical vehicle suspension systems such as MacPherson strut, double wishbone, trailing arm and solid axle suspension systems are modeled by suspension composite joints. Simple kinematic joints , force elements, tire forces and constraints are derived for the simulation of general multibody systems as well as vehicles. Equations of motions are formulated using Kane's equation with relative joint coordinates for efficient computation.
Using the developed suspension composite joints, a multibody vehicle dynamic model is formed and simulations are performed. Accuracy of the simulation results is compared to the real vehicle field test results as well as to the simulation results with a full multibody vehicle model. A double lane change, a continuous sinusoidal steering and a step steering test were performed to validate the proposed models. It is shown that the simulation results are in good agreements and coincide well with the results of the field test and simulation with the full multibody vehicle model. Thus, it is verified that the proposed vehicle models can generate very accurate predictions of real vehicle motions. With the vehicle model, real-time vehicle dynamic simulations are performed. It is found that real-time simulation is achieved on a computer with single PowerPC 604 333MHz processor with 1 millisecond integration step size.
To reduce computation time more from this results, two methods are proposed. One is using improved vehicle modeling methods and the other is exploiting sparsity of equations of motions. Three vehicle modeling methods are tested using the joints. They are kinematic steering, compliant tie-rod and force driving steer vehicle models. The last two are constructed to represent the steering compliance characteristics of a real vehicle. The vehicle model with the compliant tie-rod model is most efficient for real-time vehicle dynamic simulation and computation time is 14 per cent less than that of the kinematic steering vehicle model with the same suspension composite joints.
The coefficient matrix of linear acceleration equations has a large number of zero entries according to the relative joint coordinate numbering of a vehicle model. Two methods, sparse Cholesky and recursive block mass matrix method, are used to exploit the sparsity of the coefficient matrix. The structure of matrix can be optimized by appropriately numbering the relative joint coordinates and the methods can be used to solve the matrix exploiting its sparsity, to reduce computation time. The proposed method proved to be more efficient than the classical approach with zero calculation at least 18 per cent of total computation time for real-time vehicle dynamic simulation.
It can be concluded that the proposed vehicle models with the suspension composite joints are reliable and can achieve real-time simulation on a single processor computer with less than 1 millisecond integration step size. Thus, it can be widely used in developing highly accurate and very fast low cost automotive hardware-in-the-loop systems and high fidelity driving simulators.
