Abstract: To study the thermal expansion property of long fibre reinforced unidirectional (UD) composite, a new method was developed to create fibre random distribution models. In all the models, the fibre distribution state and the real periodic boundary condition have been taken into account. For models with high fibre volume fraction, a method called RDM (random disturbing method) was developed, which can generate models with fibre volume fraction not less than 65%. Random models generated by RDM were used to forecast the thermal expansion coefficient of M40J/TDE-85, and the thermal expansion coefficients of this composite were tested with high precision. The predicted results agree well with the experimental data, which shows that this random model can be used to predict the thermal expansion coefficients of UD composite correctly. It is rapid and precise to adopt this random model to predict the thermal expansion property of composite, and it would be convenient for materials study and engineering application.