Foreign object damage (FOD) is a big problem in aviation maintenance industry that reduces the level of safety for an aircraft. Basically, FOD is known as foreign object (FO) that can cause severity and destruction to the aircraft such as engine failure and loss of human life. Nowadays there is no FOD detection system that can classify the type of FOD that is detected optimally. This study proposes the FOD classification approach using Deep Learning. The Deep Learning method that has significant results in image recognition is the Convolutional neural network (CNN). The purpose of this study is to determine the CNN algorithm model that has optimal accuracy in the FOD classification. Convolutional Neural Network (CNN) is a multilayer neural network with a supervised learning architecture that consists of two parts, namely feature reviewers and classifiers that can be trained. The research began by designing a dataset and CNN algorithm model, then proceed with Training and Testing data. Based on the research, CNN model that has optimal accuracy using a 64 x 64 input image, learning rate value of 0.001, filter size 3 x 3, number of epochs 100, number of training data of 1200 data and test data of 20 data. The number of convolution layers used is 1 layer. The algorithm created is able to classify FOD objects into 6 classes with 86.6% accuracy and provide the best classification results into 4 classes with an accuracy of up to 90%.