In order to ensure safe travel, railway vehicles must be stable under every condition along the track. Thus, a vehicle can be certified for operation only when it can fulfil certain criteria related to the ride stability. The stability of the vehicle is highly dependent on the wheel-rail friction coefficient: higher friction results in worse ride. So, to ensure a good evaluation of the stability, the friction should be high enough during tests. The same applies to the risk of wheel flange climbing. At the present time, the wheel-rail friction can not be measured directly but there are different procedures utilized to ensure that the conditions are suitable for testing the stability of the vehicle. In this study an algorithm is proposed to estimate the wheel-rail friction coefficient by using quantities which can be measured in reality. The algorithm is tried out in computer simulations. The algorithm has two parts: in Part 1 the friction coefficient is proposed to be equal to the ratio of the total creep force divided by the normal force; in Part 2 the total creep and spin creep are estimated to observe their correlation to the estimated friction. The contact angle in Part 1 is estimated by a contact point function. In the simulations, different conditions are tried. There are four horizontal radii: tangent track, R1300m, R1000m, and R400m. Three friction coefficients are used: 0.5, 0.4 and 0.3. In addition to this, track irregularities are included. A single vehicle is simulated in two modes: capable and incapable of passive radial steering. The track irregularities caused high values of the proposed estimated friction coefficient. The values in some instances were close or equal to the input friction coefficient of the simulation. Thus, if the highest values of the estimated friction were taken over a certain distance or time, the friction of the simulation could be approximated. In most cases, the total creep was following the trend of the estimated friction. The total creep and spin creep were used as a quality factor to determine how close the estimated friction was to the simulation’s friction. In this study when the total creep was greater than 0.006 and the spin creep was less than 1.0 m-1, the estimated friction was close to the input friction. The closeness was dependent on the simulation’s friction. Higher input friction resulted in larger deviation compared to lower friction. A sensitivity analysis has been performed by deliberately introducing errors in the position of the contact point and the angle of attack. The analysis shows that the errors are not critical when the contact point is close to the tread circle. When the contact point is close to the flange, a good measurement of the wheel profile and the contact point position required to obtain accurate results. On the other hand, the errors affect the friction estimate for high spin and low total creepage. These results are discarded by the algorithm, the influence of the errors is minimized.