Figure 1 presents the experimental specimens used in the study, which consisted of ELCBs obtained from five manufacturers: A, B, C, D, and E. All ELCBs featured a rated sensitivity current of 30 mA, a rated breaking current of 2.5 kA, an AC rated voltage of 220 V, and a rated current of 20 A.

To identify data representing the highest risk, each experiment was conducted at least five times under controlled environmental conditions: a room temperature of 22 ± 5 °C and a humidity of 50 ± 5%. Figure 2 illustrates the experimental setup. To examine the operational characteristics of ELCBs under low-current series arcs, a wire connected to the secondary terminal of the test specimen ELCB was fixed at 6.7 cm using an HFIX 2.5 SQ wire. Meanwhile, a wire connecting the terminal block (TB) to a vibration tester was set at 3.2 cm to simulate a short-circuit condition. To generate series arcs, accidental short-circuit conditions were replicated by removing 0.5 cm of the sheath from the contacting parts, ensuring direct copper conductor contact. The conductors were then overlapped and separated by 0.2 cm. Vibration was applied to induce a consistent series arc caused by contact failure. To apply vibration, the vibration tester shown in Figure 3(a) was used, operating at 850 rpm determined to be the optimal setting for arc generation based on preliminary tests. Currents of 5 A, 10 A, 15 A, and 20 A, within the normal operating range of ELCBs, were applied using the AC load tester shown in Figure 3(b). Additionally, to examine operational characteristics under overload conditions, a 30 A current 1.5 times the rated current (20 A) was applied. To assess the fire risk associated with series arcs and the securing of ELCB wires, grooves were created in a plasterboard to hold the wires in place. Furthermore, 0.1 g of sawdust was added to evaluate the potential fire hazard posed by low-current series arcs. The shoulder waveform generated during series arc occurrences was observed by measuring voltage and current with an oscilloscope (Wavesurfer 64Xs-A, Lecroy Co., USA). Additionally, the thermal properties of the contact points where series arcs occurred were recorded every 5 s using a thermal imaging camera (Testo 890, Testo Co., Germany). The experiment lasted for 300 s to ensure the stability of the AC load tester. For safety reasons, the ELCBs were manually stopped if sawdust ignition was observed or if the ELCBs activated during testing.

Figure 4 presents the experimental results for a load current of 5 A. In the normal state, the wire temperature exhibited minimal variation, increasing by only 0.6 to 1.1 °C compared to the initial temperature. In contrast, during the series arc-fault state, the temperature rose significantly, increasing by a minimum of 81.4 °C and a maximum of 96.4 °C from the initial value, reaching a peak of 117.6 °C within 300 s. No ignition of the sawdust was observed. The ELCBs did not activate, as the current in both the normal and series arc-fault states remained below the rated current of the ELCBs.

Figure 5 presents the experimental results for a load current of 10 A. In the normal state, the wire temperature showed minor fluctuations, increasing by 2.1 to 2.8 °C compared to the initial temperature. However, in the series arc-fault state, the temperature increased drastically, rising by at least 314.1 °C and at most 334.8 °C from the initial value, reaching a peak of 358.0 °C within 300 s. Although this temperature exceeded the ignition threshold of sawdust, no ignition occurred. As series arcs are influenced by load currents, the thermal imaging camera detected high temperatures during arc generation. However, due to the low arc energy and a slight current reduction caused by arc resistance (R_arc) at the contact points, ignition did not occur. The ELCBs did not activate, as the current remained below the rated value in both the normal and series arc-fault states.

Figure 6 presents the experimental results for a load current of 15 A. In the normal state, the wire temperature increased by 4.0 to 4.3 °C from the initial value, a slightly higher rise than observed with a 10 A load current, though no fire risk was identified. Conversely, in the series arc-fault state, the temperature rose sharply, increasing by a minimum of 307.8 °C and a maximum of 337.0 °C and reaching a peak of 357.1 °C, which indicates a potential fire hazard. The rate of temperature increase was higher than in the 10 A case, as arc energy grew with increasing arc current. Sawdust ignition occurred between 135 and 150 s. However, the ELCBs did not activate, as the current remained below their rated value. Additionally, due to memory limitations, the thermal imaging camera captured temperature data as images every 5 s rather than as continuous video. The temperature range was set between 0 and 350 °C to more accurately monitor the initial temperature rise. While the recorded temperatures reflect conditions just before ignition, the actual temperature at the moment of flame formation was likely higher.

Figure 7 presents the experimental results for a load current of 20 A. In the normal state, the wire temperature increased slightly, rising by 12.4 to 13.0 °C compared to the initial temperature, with no fire risk identified. However, in the series arc-fault state, the temperature increased significantly, rising by at least 235.8 °C and up to 294.9 °C from the initial value, reaching a peak of 316.8 °C. At this point, sawdust ignition was observed, indicating a fire hazard. The temperature increased 4.2 times more rapidly than in the 15 A case, as arc energy further escalated with higher arc current. Sawdust ignition occurred between 30 and 35 s. However, the ELCBs did not activate, as the current remained below their rated value in both states.

Figure 8 presents the experimental results for a load current of 30 A. When overcurrent was applied, exceeding the ELCB's rated current, the wire temperature fluctuated between 24.6 °C and 41.4 °C above the initial value, reaching a peak of 65 °C under normal conditions. This temperature remains below the allowable limit of 90 °C for HFIX wires, indicating no fire risk in the wires. The ELCBs from companies A, B, and D tripped between 60 and 180 s upon detecting overcurrent, while those from companies C and E remained unaffected throughout the experiment. The lack of response from the latter ELCBs is likely due to discrepancies with the manufacturer's operating characteristic curve. In the series arc-fault state, the temperature increased significantly, rising by at least 261.5 °C and up to 297.8 °C from the initial value, reaching a peak of 321.7 °C. At this point, sawdust ignition occurred, indicating a fire risk. The rate of temperature increase was twice as high as in the 20 A case, as arc energy further escalated with the growing arc current. Sawdust ignited between 15 and 25 s. However, the ELCBs in the experimental setup did not trip before ignition occurred. For safety reasons, the ELCBs were manually turned off once the sawdust caught fire to conclude the experiment. Consequently, it was not possible to confirm whether the ELCBs would have eventually activated after ignition. If the conditions were similar to the normal state or if the current decreased slightly due to arc resistance (R_arc) at the contact point, the ELCBs might have been triggered with a slight delay.

Figure 9 presents oscilloscope waveforms for both the normal and series arc-fault states when load currents of 15 A and 20 A were applied, where the fire risk increased within the ELCB's rated current range. When a load current of 15 A was applied, the RMS value was recorded as 13.58 A in the normal state and 10.53 A in the series arc-fault state. For a load current of 20 A, the RMS value was measured at 22.43 A in the normal state and 18.19 A in the series arc-fault state. In the series arc-fault state, arc current is generated based on the load current, but it tends to decrease slightly due to the arc resistance present at the contact points. Furthermore, the shoulder phenomenon, which is characteristic of series arcs, was also observed.

Figure 10 displays images of the ELCBs in operation. Figure 10(a) shows an ELCB tripping in the normal state when an overcurrent of 30 A was applied. As shown in Figures 10(b) and 10(c), sawdust did not ignite even when a series arc occurred at currents between 5 and 10 A. In contrast, as shown in Figures 10(d) through 10(f), sawdust ignited due to a series arc at currents between 15 and 30 A, yet the ELCB did not trip. Specifically, at currents between 15 and 20 A below the ELCB’s rated threshold sawdust ignited, leading to a fire, but the ELCB failed to detect the abnormal state and did not activate. This suggests that series arcs occurring within the rated current range cannot be effectively detected or mitigated, even with ELCBs installed.