Abstract The aim of this research work is to develop a tool for condition monitoring of squirrel-cage motors using axial magnetic flux measurements, and to design a diagnostics system for electrical motors. The basic theory of the measurements and systems was found through literature reviews and was further developed from the experimental results of this research work. Fluxgate magnetometers and Hall effect sensors are not reliable enough for condition monitoring purposes, but measurements by flux coil sensors can reach adequate reliability. The useful frequency area of the flux coil sensor is from about 0.2 Hz to 15 kHz, an area is well applicable for condition monitoring of squirrel-cage motors. Output voltage is frequency dependent, increasing towards higher frequencies. Sufficient sensitivity is usually reached by a flux coil sensor having a diameter of about 30 cm and the number of turns of about 200. Sensitivity can be improved by increasing the diameter or number of turns of the coil. The sensor should be placed axially centred on the end of the motor, and measurements should be made with the loaded motor in steady operation. Output voltage is typically from the microvolt to millivolt level, however, installation inside the motor can increase it from tens of millivolts to some volts. The dynamic resolution requirement of measurement is about 70 dB and the highest line resolution needed for the spectrum analysis is about 3200 lines. Time base signal can be used to study rapid disturbances of flux caused by mechanical loading or switching of the frequency converter. Various motor failures cause specific variation to the frequency distribution of flux, so spectrum analysis is well applicable for condition monitoring. Reference measurement of each motor is required because stator winding factors, installation tolerances, operating conditions and mechanical load affect leakage flux. A broken rotor bar failure can be detected from the amplitude difference between the supply frequency and its rotor bar induced sideband. A broken rotor end ring failure can be detected by the amplitude difference between the slip frequency and the supply frequency. However, it was found that the stator current spectrum is a more reliable method of detecting both these rotor failures. The supply voltage asymmetry can also be evaluated by specific sidebands of axial flux. Turn to turn failure of the stator winding was most reliably detected by sidebands around the rotor slot pass frequencies. Equations for frequency converter supplied motors are the bases for similar equations, but the supply frequency is replaced by the output current frequency of the converter. The developed diagnostics system design for condition monitoring of ac motors includes stator current, flux coil, temperature, vibration, partial discharge, bearing current and voltage measurements. At the system diagnosis stage these measurable signals are divided to time base and frequency base signals and for each of them a fault indicator is determined. For flux coil measurements four fault indicators were found: rotor bar failure ratio, rotor end ring failure ratio, stator winding turn to turn ratio and supply voltage asymmetry operation ratio. With these failure indicators we determine failure location, state and cause. From this information a lifetime prediction of the motor is made. The results of this work are used to analyse flux coil measurements of squirrel-cage motors. In addition the research has led to a special application to monitor electric motors using an on-line condition monitoring system for paper machines and power plants.