A large-scale group Success Likelihood Index Method to estimate human error probabilities in the railway driving process

Success Likelihood Index Method (SLIM) is widely used to estimate human error probability. However, the traditional SLIM relies on a small number of experts, which may cause biases and undermine its effectiveness. To overcome this drawback, a large number of experts participate to provide fuzzy ratings and negotiate to reduce biases, which forms a large-scale group SLIM (LG-SLIM) problem. The LG-SLIM model is used to negotiate among experts to reach a unanimous agreement of the group. We improve the consensus model by constructing two mathematical programming models that provide the recommendations of both optimum preference adjustments and weight adjustment for the non-cooperative subgroup, which establishes an effective balance between the subjectivities and compromises of experts. Some typical human errors concerning target tasks are estimated in the railway driving process. To verify the proposed LG-SLIM model, the results obtained from it and individuals are compared with the simulation values from the Bayesian network respectively. The comparison shows that the results are consistent with the simulation results and conform to the experience and knowledge in railway driving. The LG-SLIM model is helpful to human error assessment when critical operations need enormous experts to obtain reliable and accurate results.