Establishing seismic fragility relationships of reinforced concrete (RC) structural walls plays an important role in the seismic risk and resilience assessment of RC structures. However, RC structural walls have different failure patterns under earthquake excitations, and different analytical models are commonly required to simulate the seismic responses of RC structural walls with different failure patterns. This brings to the difficulty of seismic fragility analysis for a wide variety of RC structural walls due to the failure-pattern-dependence. To overcome this shortcoming, a hybrid approach was developed in this study to establish the seismic fragility relationships of RC structural walls independent of failure patterns by combing experimental data and numerical simulation. The experimental data was collected from past quasi-static tests of RC structural walls having different failure patterns. The Bouc-Wen-Baber Noori (BWBN) model was used to describe the complicated hysteretic behaviors of RC structural walls under different failure patterns, whose model parameters were calibrated using the collected experimental data. Based on the calibrated BWBN models, an equivalent single-degree-of-freedom (ESDOF) system was employed to predict the seismic demand of RC structural walls. Furthermore, the predicted seismic demand of RC structural walls was compared with the seismic capacity that was both identified from the collected experimental data, developing the seismic fragility relationships of RC structural walls independent of failure pattern. For facilitating the application in practice, artificial neural network (ANN) models are finally developed to predict model parameters of the cumulative lognormal distribution for generating the seismic fragility relationships of RC structural walls rapidly. The investigation indicated that the proposed approach provides a promising way to perform the seismic fragility analysis for a wide variety of RC structural walls.