Amid the energy crisis that is threatening the global economy, the ability to forecast the production rate of crude oil for a given set of operating conditions is expedient. In this work, a feed forward back propagation neural network architecture has been designed and implemented on MatLab 7. The network was trained using field data consisting of flow tube pressure, flow line pressure, API gravity of the crude, flow temperature, base sediments and water, flow time and net production. Several network architecture ranging from one hidden layer to two hidden layers were tested with various numbers of neurons in each layer using Bayesian regularization as the training algorithm and a combination of tansignoid and pureline as transfer functions. With five network input, it was observed that the networks with a single hidden layer performed poorly, irrespective of the number of neurons in the hidden layer. The performance goal of 1e-4 was not met even when the number of epochs was raised to 600, for all networks without a second hidden layer that were trained in this work. A three layer neural network model consisting of twenty neurons in the first hidden layer, five neurons in the second hidden layer and one output layer was found to adequately fit the training data with a correlation coefficient of 0.952.Network validation output were in very good agreement with retained data.