Vehicle collision on bridges is an important issue for the transportation infrastructure management. This study explores the significance of bridge monitoring and the benefits of employing machine learning (ML) techniques to detect and classify vehicle-deck collisions on railway bridges. The ultimate goal is to transition from traditional bridge monitoring methods to a real-time monitoring system based on a ML approach, aiming to improve efficiency and accuracy in detecting bridge issues. Multiple supervised ML algorithms are evaluated to identify the most accurate model for collision detection and signal categorization. The selected ML model employs a distributed approach, enhancing its adaptability and integration into a comprehensive monitoring system for diverse bridge structures. The dataset comprises frequency, velocity, and displacement measurements collected over a one-year monitoring period from three distinct railway bridges. Additionally, a controlled experiment was conducted to identify signal patterns associated with collisions of different energy levels. The collected data underwent rigorous processing, including data cleaning, synchronization, pattern identification, and statistical analysis, to extract relevant features. The proposed model achieved an accuracy of 100% in detecting vehicle-deck collisions on railway bridges and demonstrated high accuracy in classifying other types of signals. The model provides bridge managers with a valuable digital decision support tool that aids in evaluating bridge conditions, minimizing maintenance costs, and ensuring train user safety. Furthermore, the developed approach aids in reducing disk storage and saving energy in embedded systems, enhancing its practicality and sustainability in real-world applications.