Railway wheel wear is a critical safety and maintenance issue, which is why condition monitoring is essential to prevent accidents and unnecessary delays. This work proposes a method, based on Support Vector Machines and Principal Component Analysis, to detect and classify the progression of wheel wear using mechanical vibrations. Experiments were conducted on a scaled vehicle instrumented with accelerometers and gyroscopes, which was subjected to accelerated wear tests through start and brake cycles. From the acquired signals, time and frequency domain features are extracted to obtain the dataset used to train the machine learning classification model. A support vector machine (SVM) classifier is used to identify the level of wheel wear. A hybrid SVM-principal component analysis (PCA) model is also evaluated, which is a more simplified and faster version, although with lower accuracy. Both models are evaluated and discussed. It is demonstrated that the SVM is capable of classifying different wear levels with high accuracy. The use of PCA significantly reduces the computational cost with a slight decrease in accuracy, showing a clear trade-off balance.