Air down-the-hole ( DTH) hammer drilling has long been recognized to have the potential of drilling faster than conventional rotary drill，especially in some hard rocks such as granite，sandstone，limestone，dolomite，etc． with the same weight on bit ( WOB) and rotations per minute ( ＲPM) ． So，it has been widely used in many drilling fields including mineral resource exploration drilling，oil and gas drilling and geothermal drilling． In order to reduce drilling cost by selecting optimal drilling parameters，rate of penetration ( ＲOP) should be estimated accurately and the effects of different factors on ＲOP should be analyzed． In this research，ANN model with several multi-layer perception back propagation ( BP) networks for predicting ＲOP of air DTH hammer drilling was developed using controllable parameters such as impact energy，impact frequency，WOB， ＲPM and bit operating time for the formations with a certain drillability index of rock． Several BP neural networks with the different neurons in hidden layers were developed and compared for selecting optimal architecture of ANN． The effects of the drilling parameters such as impact energy，impacting frequency，WOB，ＲPM and bit operating time on the ＲOP of air DTH hammer drilling were investigated by trained ANN． From the analyses，the optimum range of drilling parameters for providing high ＲOP were determined and analyzed for a formation with a certain drillability index of rock． The methodology proposed in this study can be used in many mathematical problems for optimization of drilling process with air DTH hammer．