With the development of the Internet of Things (IoT), an enormous amount of data is collected by the widespread use of sensors in intelligence applications, particularly in the area of connected vehicles. A connected car is a vehicle that is equipped with internet connectivity. It gathers data continuously for real-time scenario analysis across various domains using a series of onboard sensors and systems. It is may composed of GPS navigation data, speed data, acceleration data, camera data, and so on. To better detect real-time events like car crashes, a concept of the digital twin for connected vehicles is designed to show a virtual representation of the current state of the vehicles, this allows real-time monitoring of a vehicle’s status and performance. Before the data is integrated into the digital twin’s structure, the raw data from connected cars should be first preprocessed to extract meaningful feature values for analysis. However, few amount of literature has been conducted on data preprocessing for digital twins. Therefore, a concept for data preprocessing for digital twins based on streaming systems is proposed in this thesis which is specific to car crash scenarios. The research mainly works with velocity, acceleration, angular velocity, and location information due to these data having significant changes when a car happens to crash. A prototypical implementation of the concept demonstrates how the raw data generated from the Carla simulator are processed and analyzed step by step in the stream processing system with Kafka and Flink, and finally achieve the visualization of this digital twin to show the detected car crash in Grafana.