This study concerns the safety analysis of human-machine systems. Accidents are mainly caused by human operators' failure in complying with the equipment's safety rules. The aim of safety analysis was to evaluate the potential risks of a system. This analysis is usually based on a feedback from field experience describing some events unwanted already occurred in the past for similar systems. These events can have origins from technical failures or from human operator behavior. This feedback from field experience is an important input for risk analysis. In some industrial domains, the feedback from field experience is not sufficiently complete and exhaustive. This paper proposes a methodology for generating realistic unwanted human behaviors. This study aimed to anticipate the possible behaviors of the human operators that do not respect the safety rules to accomplish a given task. This paper proposes an alternative action plan generator (A2PG), which predicts the alternatives for the probable human action plans in which the human operators do not respect the safety rules. Based on a formalism used for task scheduling, the specifications of A2PG are presented. The strategy of generating the alternative action plans taking into account the failure to comply with the safety rules is also presented. Finally, a case study coming from agricultural domain with a riding lawn mower is presented in order to evaluate the application and the relevance of the proposed tool;