Magnetic Fields and Induced Power in the Induction Heating of Aluminium Billets

Abstract: Induction heating is a common industrial process used for the reheating of billets before extrusion or forging. In this work the influence of the coil and work piece geometry, the effect of the electrical properties of the work piece, and the coil current and frequency, on the magnetic flux density and resulting work piece heating rates were studied. A combination of 1D analytical solutions, 2D axial symmetric finite element modelling and precise measurements has been used.Dozens of heating and magnetic field experiments have been conducted, with steadily increasing sophistication and measurement accuracy. The development of the experimental techniques will be described in the ‘cover’ and related to the later results published in the supplements. Experimental results are compared to predictions obtained from analytical and numerical models. The published measurements obtained for the billet heating experiments consisted of: billet electrical conductivity with <0.5% error, applied currents with <1% error, magnetic flux densities with 1-2% error, calorifically determined heating rates with <2% error and electrical reactive power with <~2% error. 2 D axial symmetric finite element models were obtained, which describe the measured results with less than a 2% difference (i.e. an ‘error’ of the same magnitude as the measurement uncertainty). Heating and reactive power results predicted by the FEM model are in excellent agreement with analytical solutions from 50 Hz to 500 kHz (differences from <1% to 6%).A modified 1D short coil correction factor is presented which accounts for the interaction of the coil and work piece geometry, electrical properties and operating frequency, on the average magnetic flux density of the coil/work piece air-gap and the resulting heating rate. Using this factor, the average magnetic flux density in the air-gap can be estimated analytically within 2-3% and the heating rates of billets of known electrical properties can be estimated, with typical errors on the order of 5%.