Present study reports acoustic emission (AE) technique to evaluate corrosion damages of bottom plates (floor conditions) in above-ground tanks. Artificial AE signals were generated by pencil lead breaks at arbitrary locations on the bottom of a tank (300 kL in capacity) to investigate accuracy of AE source location. Attenuation of AE waves was shown to be very small in liquid. AE source location analysis was conducted, using a neural network (NN). Input and output units of the NN were arrival time differences between four AE sensors and coordinate of the AE source location, respectively. Arrival time differences of AE waves were determined by visual observation of the first signal arrivals and threshold crossing times of the AE signal normalized by its peak amplitude. It was concluded that accurate AE source location can be obtained by the decision process resulted from automated readings of threshold crossing time, based on the NN trained method by theoretical calculation.