In countries with rough weather conditions, frequent delays cause railway companies to waste time and money. Many of these delays are related to the train detection systems, as the old DC track circuits are still used in some countries, including Sweden, our case study. Since the most important factor in the railway system is safety, in some cases, the train detection system gets incorrect information and detects a non-existent train. The train slows down to avoid a problem in the track (with other trains or other faults), causing prolonged delays with cascading effects. The analysis in this licentiate contributes to the detection and reduction of TC failures; this, in turn, will save money for the railway community.A classification of the most probable causes of failures related to the train detection system was derived from the Swedish failures report database 0FELIA. After classifying failures, we focussed on the three most common worst case scenarios: low resistance between the rails, external interference such as a lightning strike, and iron-powder-bridges in the insulated joint.Electromagnetic interferences (EMI) are a problem for the railway system in general. One source of electromagnetic (EM) transients is the return current harmonic produced by the engine of the rolling stock itself. In the first stage of this licentiate, we implemented a Matlab model of the power supply system of the Swedish railway infrastructure, using the characteristics and previous measures of a real source. A model of a train as an active load validated by the manufacturer was integrated as a subsystem in different positions of the infrastructure. This method was used to study the behaviour of the low frequency system from an electrical point of view but it could also be used as input for an electromagnetic model using high frequencies. The model was validated through measurements taken in northern Sweden.In addition, a 3D model of the whole railway system was proposed. The simulation software was CST STUDIO SUITE® (Computer Simulation Technology Studio Suite), supported by real measurements on site and the lab to tune and validate the model. The results of the simulation show that the model fits with reality and is reliable for the study of track circuit sections.Some measurements followed the current standards, but we also analysed points not covered by them, allowing us to update the current standards