The aim of this paper is to propose a method for detecting the occurrence of wheel flats and estimating their size from data of multiple wayside train monitoring systems during normal operation. Thereby, the need for manual inspections can be reduced or at least becomes more efficient, releasing resources to focus on other aspects in maintenance. Moreover, decision making on wheelset exchanges and wagon shunting is improved and the risk for human-errors is reduced. The detection of wheel flats is based on the fact that wheel flats usually occur on both wheels of a wheelset at the same time rendering a jump in the recorded rail-wheel contact forces by wheel impact load detectors. The wheel flat length estimation uses a model-based approach where simulated forces are matched with the measured forces by adjusting the flat length in the simulation. The scheme is evaluated on several real-life cases showing promising result, and will be further validated in field tests.