Experimental methods

We constructed our EPM concept (Fig. 3) as follows. The yoke was made of M270-35A [41] backlack sheet metal. We used a Bakker Magnetics N45-grade PM. The four windings with 4 × 240 turns of wire with a diameter of 3.2 mm were powered via an adjustable DC power supply (Velleman LABPS3020). To achieve a magnetic flux density of 1 T inside the air gap, an electric current of approximately 3.75 A had to flow through each winding, which contributed 0.5 T; the other 0.5 T was contributed by the PM, cumulatively achieving 1 T inside the air gap. Simultaneously, to achieve 0 T inside the other air gap, an electric current of approximately 3.75 A had to flow in the opposite direction through the windings, which negated the 0.5 T provided by the PM. This is explained in detail in Appendix A. The time variation in magnetic flux density inside the gap was regulated by transistors, which were controlled by an Arduino Mega. The measurements were conducted by a Hall probe (Magnet-Physik Gauss-/Teslameter FH 54) measuring the magnetic flux density in the air gap, current clamps (Chauvin Arnoux AC/DC 10 A/1 V 100 A/1 V E3N) measuring the electric current through the EM winding, and a Keysight EDUX1002A oscilloscope. The experiments were repeated at frequencies of 1, 2.5, 5, 10, 15, 25 and 50 Hz. A more detailed description of the experiment (magnetic energy recovery, power input and energy efficiency) and measuring uncertainties are given in Appendices D and E. The electronics (Appendix A) contains transistors with resistance between drain and source in the ˝on˝ state R_DS,on = 38 mΩ and diodes with forward voltage U_f = 1.25 V. The temperature of the EPM during operation was measured using Type-K thermocouples (0.2 mm diameter), the National Instruments NI-CDAQ-9174 data acquisition system and an NI 9214 temperature module. Four thermocouples were placed on the yoke’s surface (two on the middle pillar and two on the outer pillar), and two thermocouples were placed on each winding surface. The expanded measurement uncertainty was ±0.3 K.

Proposed EPM design. a Constructed proof-of-concept (the Arduino Mega 2560 is also presented for scale). b 3D model of the proposed EPM design.