Online Identification of Vehicle Dynamic Systems for Condition Monitoring

This PhD project focuses on developing online monitoring approaches for suspension
systems based on vibration analysis, aiming at guaranteeing the safe and efficient operations
of vehicles including the railway and autonomous vehicles.

Based on Operational Modal Analysis (OMA), which has been proven more effective in the
field of structural health monitoring, a novel OMA algorithm, entitled the Correlation Signal
Subset-based Stochastic Subspace Identification (CoSS-SSI), is proposed in this thesis to
identify the inherent vibration modes of a car body and railway bogie frame to assess the
health of the vehicle suspension system. The proposed novel OMA method is developed in
the knowledge that the basic framework of SSI makes it applicable to nonlinear systems with
nonstationary responses in the presence of high noise levels.

For the CoSS-SSI method, the measured raw signals are divided into short segments; then
the correlation function of each data segment is calculated, which performs the first noise
reduction. After that, the obtained correlation function for the segments are divided into
subsets according to their minimum amplitudes, and then each subset is averaged to further
reduce the noise. Different correlation signal subsets can reduce nonlinear effects such as
high damping, on OMA. Lastly, each subset of the averaged correlation signals is utilised to
accurately identify the modal parameters based on SSI.

A 3-DOF vibration system was developed in an initial simulation study developed to
evaluate the performance of CoSS-SSI, which showed that CoSS-SSI was superior to other
conventional OMA methods like Cov-SSI in extracting useful modal information on system
behaviour. In addition, a quarter vertical vehicle model was constructed to investigate the
effects of periodic pulses and harmonics on OMA. It was found that periodic pulses have no
impacts on OMA, but harmonics can cause significant adverse effect. Then, cepstrum editing
was introduced to eliminate the harmonic effect on OMA, and its performance was first
verified by simulated data and later by experimental data obtained from full-scale rig tests.

Experimental studies were carried out to verify the feasibility of applying the proposed
method for the online monitoring of suspension systems. In the first set of experiments,
accelerometers were installed at the four corners of a car body and it was shown that with
CoSS-SSI these comprised a robust and cost-efficient system for monitoring the suspension
system. These results were confirmed by using CoSS-SSI to identify the modal parameters
of a road vehicle suspension using the measured vibrations of a real car running normally on
a traditional country road near Huddersfield, UK. These experiments confirmed that CoSSSSI
had the capability to extract the inherent vibration modes of the vehicle suspension
system.

Importantly, a 1/5th scale roller rig and then an Y25 bogie were employed to verify the
potential of CoSS-SSI for railway vehicle suspension monitoring. The outcomes from roller
rig experiments showed that the novel CoSS-SSI proposed here is also feasible for the
successful online monitoring of railway vehicle suspension systems.