An alarm system forms an essential part of the operator interfaces for automatically monitoring the plant state deviations and attracting the plant operator's attention to changes that require his/her intervention. To design an efficient alarm system that helps operators quickly identify the reason of abnormality and response to the situation is a vital work for the safe operation of the plant. In the EEMUA 191 Guide pressed by Engineering Equipment and Materials Users Association (EEMUA), general principles of design and evaluation of plant alarm systems are proposed. Based on this guide, we propose a roadmap about the plant alarm system design, which includes a circle of design, evaluation, alarm system rationalization and reevaluation. In the basic design stage, based on cause-effect analysis, a method is proposed for choosing a small number of alarm source signals but with more information for fault identification. In the alarm system rationalization stage, we propose a repeating alarm suppression algorithm and improve the alarm display interface. As for the evaluation and reevaluation of the designed alarm system, we refer to the criteria provided by EEMUA 191 Guide, and introduce other indexes such as the malfunction identification time of a human operator model as well. The design methods are applied to the alarm system of a boiler plant simulator as a case study. Evaluation results show the usefulness of the proposed methods.