[en] In safety-critical systems, the generation failure of an actuation signal is caused by the concurrent failures of the automated systems and an operator action. These two sources of safety signals are complicatedly correlated. The failures of sensors or automated systems will cause a lack of necessary information for a human operator and result in error-forcing contexts such as the loss of corresponding alarms and indications. In the conventional analysis, the Human Error Probabilities (HEP) are estimated based on the assumption of 'normal condition of indications and alarms'. In order to construct a more realistic signal-generation failure model, we have to consider more complicated conditions in a more realistic manner. In this study, we performed two kinds of investigation for addressing this issue. We performed the analytic calculations for estimating the effect of sensors failures on the system unavailability and plant risk. For the single-parameter safety signals, the analysis result reveals that the quantification of the HEP should be performed by focusing on the 'no alarm from the automatic system and corresponding indications unavailable' situation. This study also proposes a Condition-Based Human Reliability Assessment (CBHRA) method in order to address these complicated conditions in a practical way. We apply the CBHRA method to the manual actuation of the safety features such as a reactor trip and auxiliary feedwater actuation in Korean Standard Nuclear Power Plants. In the case of conventional single HEP method, it is very hard to consider the multiple HE conditions. The merit of CBHRA is clearly shown in the application to the AFAS generation where no dominating HE condition exits. In this case, even if the HE conditions are carefully investigated, the single HEP method cannot accommodate the multiple conditions in a fault tree. On the other hand, the application result of the reactor trip in SLOCA shows that if there is a dominating condition, the use of single HEP method could be a practical way of developing a model