Pantograph-catenary conditions monitoring is a challenging issue in railway infrastructures. Thus, the failure of these critical systems causes considerable detriment with regard to availability and passenger safety and requires substantial human and financial resources for maintenance. However, rail operators still use traditional methods to control many equipment thorough visual inspections. In recent years, several research have suggested the use of Artificial Intelligence for conditions monitoring, the latest still incipient leading sometimes to unwarranted investments. In this context, the paper examines recent advancements in pantograph condition monitoring over the past five years, within the domain of predictive maintenance. It highlights the methods, algorithms, and monitored parameters through a comprehensive analysis of prior studies. The analysis revealed that the Convolutional Neural Networks (CNNs) and image processing techniques have been extensively employed. Furthermore, a comparative evaluation is provided, emphasizing performance across key metrics such as accuracy, precision and MRSE. Finally, we have proposed an unified Predictive Maintenance workflow of wear detection based on CNN and deep learning. In conclusion the study outlines the major challenges and limitations of current practices.