In this paper, we propose a data-driven robust fault tolerant preview control scheme for discrete-time LTI systems with completely unknown dynamics. In general, this work consists of three key design parts. Firstly, the Markov parameters sequence is identified to extract system feature from data. Then, robust linear quadratic preview control policy composed of finite data feedback, integral operation and preview action is designed through establishing data-space characterisation and solving the related convex optimisation problem. Thirdly, an active fault tolerant method composed of fault detection, isolation, estimation and corresponding compensation is further constructed to accommodate fault influence. Note that all these involved designs are performed at Markov parameters sequence level. Thus above integrated preview control scheme can avoid the priori knowledge requirements including system matrices, order, and state variables. The effectiveness of the obtained results is finally verified via a case study of the injection moulding machine.