Developing a fast computing second-order transfer function engine model using system identification for internet distributed hardware in the loop application

Abstract

Despite technological advancement, physical testing of components is still a necessity. Relying on simulations might not be enough especially in designing hybrid vehicles due to the complexity of the physical connection between the different components. However, assembling those components for physical testing will lead to an increase in the development cost. A good compromise between accuracy and cost is hardware in the loop simulation or internet distributed hardware in the loop. This is the aim of the virtually connected hybrid vehicle project. Realising a virtually connected hybrid vehicle and testing each component using internet distributed hardware in the loop will lead to the reduction of development cost caused by the need for physical testing and will increase the accuracy compared to the simulation. The project to realise a virtually connected hybrid vehicle is a 3-year long research study program between the institute of digital engineering and 6 top UK universities Bath, Loughborough, Warwick, Newcastle, Nottingham and University College London (UCL) with each university responsible for one component of the hybrid vehicle.
Realising an internet distributed hardware in the loop presents big research challenges, one of these challenges is guaranteeing a stable and transparent hardware in the loop simulation at each university despite adverse factors imposed by the internet connection (delay, jitter and packet loss). Understanding the influence of these factors on the stability and transparency of the locale HiL will allow the development of feasible solutions. However, the internet is not the only reason for the distortion, instability and transparency. Realising hardware in the loop simulation or power hardware in the loop comes with plenty of challenges. A literature review examining different aspects of realising hardware in the loop, power hardware in the loop and internet distributed hardware in the loop is presented.
Having understood the challenges that need to be overcome to realise a stable and accurate hardware in the loop and internet distributed hardware in the loop, communication between the different universities needs to be established. However, having different universities exchange data between themselves is not enough. A communication architecture between the different sites needs to be established. Two communication architecture will be analysed, the multi-way communication architecture and the hierarchal communication architecture. A comparison between the two communication architectures was drawn and the hierarchal communication architecture was chosen to be used in this PhD. More than that, the quality of the internet connection between the University of Bath and the universities needs to be characterised to understand the impact it will have on the experiment. Finally, knowing where to divide the system and how to break the vehicle apart is crucial seeing that the impact of the coupling point is considerable and needs to be taken as a design characteristic when possible.
Some universities included in the project will be running electrical components on their test rigs requiring high data exchange with remote sites. Exchanging data over the internet will not be feasible, so having a model that can emulate the remote component is key to ensuring stability. So, a fast-computing engine model that can emulate the behaviour of the engine running in the test rig is needed. An analytical analysis to prove that the mean value engine speed response can be described by a second-order transfer function. This theory will lead to creating an engine model with a variable second-order transfer function that emulates the engine running in the test rig. That theory will be put to the test by being performed in the simulation environment using an engine model developed by AMESIM.
The AMESIM engine model was used to identify the impact of the delay on the transparency of the engine behaviour. That was done by generating a simplified model that emulates the overall vehicle and what would be expected to be received by the engine site. Two coupling points will be considered and full analysis for different delays will be performed. More than that, the AMESIM engine model will be used to test the theory of generating an engine model that emulates the engine speed response using system identification assuming that the engine behaviour can be described by a second-order transfer function. The identification process will be discussed in detail and the variable transfer function engine model generated will be compared to the AMESIM engine when tested on different drive cycles.
After being tested and proven in the simulation environment, the process will be repeated experimentally using a Puma 2.2 litre ford diesel engine. The identification process will be performed on the engine to create a model of that engine, then the engine will run on a drive cycle. The variable transfer function engine model generated using system identification will be run on the same drive cycle to validate the accuracy of the modelling technique introduced. Having found a discrepancy between the results in the simulation and the ones in the experiment, a closer analysis of the data collected in the experiment was performed. Having found that the experimental conditions were different than the ones in the simulation, the conditions encountered experimentally were replicated in the simulation. Once the AMESIM engine was subject to similar experimental conditions, the results were closer to the one found in the experiment. Seeing that the reason behind that difference was determined, the experiment was repeated ensuring that the conditions encountered in the previous experiment were eliminated. This led to results more in line with the ideal simulation case and proved that the modelling technique introduced of assuming the engine speed response can be described by a variable second-order transfer function model is accurate.

Date of Award	13 Dec 2021
Original language	English
Awarding Institution	University of Bath
Supervisor	Sam Akehurst (Supervisor), Richard Burke (Supervisor) & Chris Brace (Supervisor)