Date of Award


Publication Type

Master Thesis

Degree Name



Mechanical, Automotive, and Materials Engineering

First Advisor

Johrendt, J.


control, co-simulation, neural, Semi-Active, sliding, suspension




Several FCA vehicles are fitted with semi-active damper systems which modulate the level of damping implemented in the vehicle suspension system to improve both the handling and ride quality felt by vehicle’s occupants. Durability simulations are necessary to analyze a vehicle’s or a component’s structural integrity over an expected lifespan. Performing durability simulations in a virtual environment has streamlined the traditional development cycle by reducing the need to construct physical prototypes and conduct physical road or bench tests. It is essential that the vehicle is modeled as accurately as possible in the virtual environment to ensure the results are representative of real-world performance. Presently, the incorporation of a semi-active damper system in a virtual durability simulation involves the expensive and resource intensive use of empirically obtained data. The goal of this project is to improve the fidelity and efficiency of durability simulations by including the loading effects of a semi-active suspension system. To accomplish this, several semi active suspension control algorithms and practical considerations are studied. Using a car model developed in Simulink©, a neural network, clipped optimal control, and sliding mode control algorithms are developed to approximate operating characteristics of the supplier controller. The development of each controller, along with appropriate tuning and validation procedures in Simulink©, are presented. A process known as co-simulation is then used to integrate each of the chosen semi-active damper control systems into durability simulations used in vehicle development processes at FCA. Co-simulation is a process wherein the controller is executed in parallel with MSC Adams© CAE durability simulation software using Matlab©/Simulink©. The accuracy of the neural network, sliding mode controller, and clipped optimal controller are validated by correlating results to a Co-simulation carried out with a supplier controller. It is found that the performance of the neural network controller resulted in output chattering throughout the simulation. While performance is acceptable in ranges where the output data is expected to be low frequency and low amplitude, instances where this was not the case induced chattering events. These events are most likely due to the neural network receiving inputs outside of the range of data which it was trained on.