Type-I X-ray bursts are recurring thermonuclear explosions on the surface of accreting neutron stars. Matching observed bursts to computational models can help to constrain system properties, such as the neutron star mass and radius, crustal heating rates, and the accreted fuel composition, but systematic parameter studies to date have been limited. We apply Markov chain Monte Carlo methods to 1D burst models for the first time, and obtain system parameter estimations for the `Clocked Burster', GS 1826$-$238, by fitting multiple observed epochs simultaneously. We explore multiple parameters which are often held constant, including the neutron star mass, crustal heating rate, and hydrogen composition. To improve the computational efficiency, we precompute a grid of 3840 KEPLER models - the largest set of 1D burst simulations to date - and by interpolating over the model grid, we can rapidly sample burst predictions. We obtain estimates for a CNO metallicity of $Z_\mathrm{CNO} = 0.010^{+0.005}_{-0.004}$, a hydrogen fraction of $X_0 = 0.74^{+0.02}_{-0.03}$, a distance of $d \sqrt{\xi_\mathrm{b}} = 6.5^{+0.4}_{-0.6}\, \mathrm{kpc}$, and a system inclination of $i = {69^{+2}_{-3}}^{\circ}$.

In low-mass X-ray binaries, the accretion of stellar material onto a neutron star can fuel unstable thermonuclear flashes known as Type I X-ray bursts. Simulating these events using computational models can provide valuable information about the nature of the accreting system. One-dimensional (1D) astrophysics codes with large nuclear reaction networks are the current state-of-the-art for simulating X-ray bursts. These codes can track the evolution of isotopes through thousands of nuclear reaction pathways, to predict the released nuclear energy and final composition of the ashes. In this thesis, I make extensive use of KEPLER, a 1D code at the forefront of these efforts. I first present improvements to the setup and analysis of KEPLER burst models. By accounting for nuclear heating in the initial conditions, I shorten the thermal burn-in time, thereby reducing computational expense and producing more consistent burst trains. To model bursts fueled by transient accretion events, I perform the first such simulations with fully time-dependent accretion rates. Building upon previous efforts to model the "Clocked Burster", GS 1826$-$238, I precompute a grid of 3840 simulations and sample the interpolated results using Markov Chain Monte Carlo (MCMC) methods. By comparing the predictions to multi-epoch observations, I obtain posterior probability distributions for the system parameters. I then extend these MCMC methods to the pure-helium burster, 4U 1820$-$30, using a grid of 168 simulations. Finally, I discuss potential improvements for future studies, to further develop the computational modelling of accreting neutron stars.

We present a new method of matching observations of Type I (thermonuclear) X-ray bursts with models, comparing the predictions of a semi-analytic ignition model with X-ray observations of the accretion-powered millisecond pulsar SAX J1808.4--3658 in outburst. We used a Bayesian analysis approach to marginalise over the parameters of interest and determine parameters such as fuel composition, distance/anisotropy factors, neutron star mass and neutron star radius. Our study includes a treatment of the system inclination effects, inferring that the rotation axis of the system is inclined $\left(69^{+4}_{-2}\right)^\circ$ from the observers line of sight, assuming a flat disc model. This method can be applied to any accreting source that exhibits Type I X-ray bursts. We find a hydrogen mass fraction of $0.57^{+0.13}_{-0.14}$ and CNO metallicity of $0.013^{+0.006}_{-0.004}$ for the accreted fuel is required by the model to match the observed burst energies, for a distance to the source of $3.3^{+0.3}_{-0.2}\,\mathrm{kpc}$. We infer a neutron star mass of $1.5^{+0.6}_{-0.3}\,\mathrm{M}_{\odot}$ and radius of $11.8^{+1.3}_{-0.9}\,\mathrm{km}$ for a surface gravity of $1.9^{+0.7}_{-0.4}\times10^{14}\,\mathrm{cm}\,\mathrm{s}^{-2}$ for SAX J1808.4--3658.

Many distinct classes of high-energy variability have been observed in astrophysical sources, on a range of timescales. The widest range (spanning microseconds-decades) is found in accreting, stellar-mass compact objects, including neutron stars and black holes. Neutron stars are of particular observational interest, as they exhibit surface effects giving rise to phenomena (thermonuclear bursts and pulsations) not seen in black holes. Here we briefly review the present understanding of thermonuclear (type-I) X-ray bursts. These events are powered by an extensive chain of nuclear reactions, which are in many cases unique to these environments. Thermonuclear bursts have been exploited over the last few years as an avenue to measure the neutron star mass and radius, although the contribution of systematic errors to these measurements remains contentious. We describe recent efforts to better match burst models to observations, with a view to resolving some of the astrophysical uncertainties related to these events. These efforts have good prospects for providing complementary information to nuclear experiments.

Modelling of thermonuclear X-ray bursts on accreting neutron stars has to date focused on stable accretion rates. However, bursts are also observed during episodes of transient accretion. During such events, the accretion rate can evolve significantly between bursts, and this regime provides a unique test for burst models. The accretion-powered millisecond pulsar SAX J1808.4-3658 exhibits accretion outbursts every 2-3 years. During the well-sampled month-long outburst of 2002 October, four helium-rich X-ray bursts were observed. Using this event as a test case, we present the first multi-zone simulations of X-ray bursts under a time-dependent accretion rate. We investigate the effect of using a time-dependent accretion rate in comparison to constant, averaged rates. Initial results suggest that using a constant, average accretion rate between bursts may underestimate the recurrence time when the accretion rate is decreasing, and overestimate it when the accretion rate is increasing. Our model, with an accreted hydrogen fraction of $X=0.44$ and a CNO metallicity of $Z_\mathrm{CNO}=0.02$, reproduces the observed burst arrival times and fluences with root mean square (RMS) errors of $2.8\,\mathrm{h}$, and $0.11\times 10^{-6}\,\mathrm{erg\, cm^{-2}}$, respectively. Our results support previous modelling that predicted two unobserved bursts, and indicate that additional bursts were also missed by observations.