We propose a novel phase association and source scanning method by incorporating the norm‐grid search and a beam search technique together to associate seismic phases, and to determine the accurate location of a particular event simultaneously. The method is called the beam search‐based phase association and source scanning (BSPASS). Using synthetic travel times for two seismic networks with different station spacings, we examine the practicability and accuracy of the BSPASS method. The results show that BSPASS could constrain event location with high precisions, if the standard deviation of travel‐time errors is less than 0.4 s. The method works well even for those earthquakes located at the network boundary. We applied the BSPASS method to the beginning day of the 2018 Purcell mountains earthquake swarm. Two times more events are detected automatically, and similar spatiotemporal distribution of these earthquakes is recognized by our method compared to the U.S. Geological Survey catalog. We also investigate the flexibility and robustness of the BSPASS method by adding Gaussian noises to phase arrivals. With the increasing of noise levels, the location errors grow accordingly. Based on our tests, we suggest that there might be an upper limit of phase picking errors for a particular source array configuration, for example, 0.4 s in this study, above which the global optimization procedure fails to resolve a better source location under the constraint of travel‐time residuals than that of the norm‐grid searching.