In past decades, impact hammers have played a key role in underground construction. Simply in Istanbul, impact hammers have been used to excavate more than 20 km of metro tunnels. Thus, determining the instantaneous breaking rate (IBR) of an impact hammer is attracting increasing attention. A number of IBR prediction models have been developed for impact hammers. However, there is still a demand for models that require a smaller number of easy-to-obtain rock properties as inputs and provide a reasonable level of accuracy, a wide range of applications, and high reliability. This study had the goal of developing such a prediction model based on an investigation of two subway tunnels built in Istanbul. In order to enhance the results generated by multiple linear regression analysis, a customized tool was developed for the non-linear analysis of a relatively large set of data collected for the present research. Gene expression programming (GEP) and particle swarm optimization (PSO) were merged to create a non-linear analysis tool. The GEP-PSO algorithm was trained using 80% of the available data, with the remaining 20% reserved to validate the results. The output of the algorithm was presented in the form of a mathematical equation that predicted the IBR using the uniaxial compressive strength, rock quality designation, Schmidt hammer rebound value, and machine power as input parameters. The predicted IBR values were in remarkable agreement with the recorded values. In order to verify the efficiency of the proposed prediction model, it was successfully tested against previously developed models for which input parameters were available. In addition, the proposed model was investigated under hypothetical circumstances to ensure that it legitimately described the performance changes due to changes in the input parameters. The model developed in this research is proposed as an accurate and reliable tool for predicting the performance of impact hammers over a wide application range.